计算机科学与技术（参考文献）3-Dimensional Reconstruction on Tagged Packages via RFID Systems

3-Dimensional Reconstruction on Tagged Packages via RFID Systems Yanling Bu,Lei Xie,Jia Liu,Bingbing He,Yinyin Gong and Sanglu Lu State Key Laboratory for Novel Software Technology,Nanjing University,China yanlingbu@foxmail.com,Ixie@nju.edu.cn.jialiu.cs@gmail.com, [hebb,yygong}@dislab.nju.edu.cn,sanglu@nju.edu.cn Abstract-Nowadays,3D reconstruction has been introduced in monitoring the package placement in logistic industry-related applications.Existing 3D Package reconstruction methods are mainly orientation Package based on computer vision or sensor-based approaches,which are stacking limited by the line-of-sight or battery life constraint.In this paper, we propose RF-3DScan to perform 3D reconstruction on tagged packages via passive RFID,by attaching multiple reference tags -Antenna onto the surface of the packages.The basic idea is that by moving the antenna along straight lines within a constrained 2-dimensional space,the antenna obtains the RF-signals of the Moving direction reference tags attached on the packages.By extracting the phase differences to build the angle profile for each tag,RF-3DScan can compare the angle profiles of the different reference tags and Fig.1.3D reconstruction on tagged packages via mobile scanning derive their relative positions,then further determine the package Existing 3D reconstruction methods are mainly based on orientation and stacking for 3D reconstruction.We implement computer vision or sensor-based approaches.Computer vision- RF-3DScan and evaluate its performance in real settings.The experiment results show that the average identification accuracy based approaches leverage the cameras to capture the appear- of the bottom face is about 92.5%,and the average estimation ance and build the 3D profiles of objects [1.21.They are able error of the rotation angle is about 4.08. to reconstruct the shape of objects in a vivid approach.The main disadvantage is the line-of-sight constraint in capturing I.INTRODUCTION images,which leads to blind angles in 3D reconstruction Nowadays,the traditional logistic industry-related appli- with only one fixed camera.Sensor-based approaches usually cations,such as warehouse management and logistic trans- leverage inertial sensors attached to the items to detect the portation,are emerging with brand new requirements.For orientation variation of the specified items [3,4].However, example,during the process of warehouse management or they suffer from high hardware cost of the sensors,as well logistic transportation,the packages are usually required to be as the limited battery life for the sensors.Fortunately,the placed according to some specified regulations.In particular,rising use of RFID technology in the logistic industry has in regard to a single package,if it contains orientation-sensitive brought brand new opportunities to 3D reconstruction on goods,such as chemical reagents or precision instruments, packaged objects.In current logistic industry,RFID tags have then it is prohibited from being rollover or upside down; been widely used to label the packages with exact logistics in regard to multiple packages,they are also required to be information.In comparison to the above two approaches,the precisely arranged in some specified order,e.g.,heavy objects passive RFID tag is battery-free and very cheap,and the are placed on the bottom,whereas light objects are placed backscatter-based communication from RFID is not limited on the top,to ensure safety in the transportation.To deal by the line of sight requirement.Most importantly,for most with the above requirements,the technology of 3-dimensional logistic applications,the RFID systems are already deployed (3D)reconstruction has been introduced to tackle these issues in the sites to scan and identify the tagged packages. in monitoring the package placement.3D reconstruction is a Therefore,in this paper,we propose RF-3DScan,which process of capturing the shape and appearance of a single or aims to perform 3D reconstruction on packaged objects via multiple real objects.In principle,there are two key aspects the RFID systems (in Fig.1).Our idea is based on the to realize 3D reconstruction for packaged objects:1)Package observation that by attaching multiple tags onto the surface orientation for a single object,it refers to determining the of the packages,we are able to derive the 3D orientation of relative orientation for each package,i.e..figuring out the each single package and the 3D stacking situation of multiple bottom/top face,as well as the angles of the other vertical sides packages according to the backscattered RF-signals from these for the specified object in the specified coordinate system.2) reference tags.Our approach of RF-3DScan is as follows:We Package stacking for multiple objects,it refers to determining attach a set of reference tags on the surface of the packages, the relative stacking situation for multiple packages,i.e.,then we utilize a single RFID antenna to continuously scan performing the relative localization of multiple objects. the tagged packages,while the antenna is moving along

3-Dimensional Reconstruction on Tagged Packages via RFID Systems Yanling Bu, Lei Xie, Jia Liu, Bingbing He, Yinyin Gong and Sanglu Lu State Key Laboratory for Novel Software Technology, Nanjing University, China yanlingbu@foxmail.com, lxie@nju.edu.cn, jialiu.cs@gmail.com, {hebb, yygong}@dislab.nju.edu.cn, sanglu@nju.edu.cn Abstract—Nowadays, 3D reconstruction has been introduced in monitoring the package placement in logistic industry-related applications. Existing 3D reconstruction methods are mainly based on computer vision or sensor-based approaches, which are limited by the line-of-sight or battery life constraint. In this paper, we propose RF-3DScan to perform 3D reconstruction on tagged packages via passive RFID, by attaching multiple reference tags onto the surface of the packages. The basic idea is that by moving the antenna along straight lines within a constrained 2-dimensional space, the antenna obtains the RF-signals of the reference tags attached on the packages. By extracting the phase differences to build the angle profile for each tag, RF-3DScan can compare the angle profiles of the different reference tags and derive their relative positions, then further determine the package orientation and stacking for 3D reconstruction. We implement RF-3DScan and evaluate its performance in real settings. The experiment results show that the average identification accuracy of the bottom face is about 92.5%, and the average estimation error of the rotation angle is about 4.08◦. I. INTRODUCTION Nowadays, the traditional logistic industry-related appli￾cations, such as warehouse management and logistic trans￾portation, are emerging with brand new requirements. For example, during the process of warehouse management or logistic transportation, the packages are usually required to be placed according to some specified regulations. In particular, in regard to a single package, if it contains orientation-sensitive goods, such as chemical reagents or precision instruments, then it is prohibited from being rollover or upside down; in regard to multiple packages, they are also required to be precisely arranged in some specified order, e.g., heavy objects are placed on the bottom, whereas light objects are placed on the top, to ensure safety in the transportation. To deal with the above requirements, the technology of 3-dimensional (3D) reconstruction has been introduced to tackle these issues in monitoring the package placement. 3D reconstruction is a process of capturing the shape and appearance of a single or multiple real objects. In principle, there are two key aspects to realize 3D reconstruction for packaged objects: 1) Package orientation for a single object, it refers to determining the relative orientation for each package, i.e., figuring out the bottom/top face, as well as the angles of the other vertical sides for the specified object in the specified coordinate system. 2) Package stacking for multiple objects, it refers to determining the relative stacking situation for multiple packages, i.e., performing the relative localization of multiple objects. 7DJ $QWHQQD 3DFNDJH RULHQWDWLRQ 3DFNDJH VWDFNLQJ Fig. 1. 3D reconstruction on tagged packages via mobile scanning Existing 3D reconstruction methods are mainly based on computer vision or sensor-based approaches. Computer vision￾based approaches leverage the cameras to capture the appear￾ance and build the 3D profiles of objects [1, 2]. They are able to reconstruct the shape of objects in a vivid approach. The main disadvantage is the line-of-sight constraint in capturing images, which leads to blind angles in 3D reconstruction with only one fixed camera. Sensor-based approaches usually leverage inertial sensors attached to the items to detect the orientation variation of the specified items [3, 4]. However, they suffer from high hardware cost of the sensors, as well as the limited battery life for the sensors. Fortunately, the rising use of RFID technology in the logistic industry has brought brand new opportunities to 3D reconstruction on packaged objects. In current logistic industry, RFID tags have been widely used to label the packages with exact logistics information. In comparison to the above two approaches, the passive RFID tag is battery-free and very cheap, and the backscatter-based communication from RFID is not limited by the line of sight requirement. Most importantly, for most logistic applications, the RFID systems are already deployed in the sites to scan and identify the tagged packages. Therefore, in this paper, we propose RF-3DScan, which aims to perform 3D reconstruction on packaged objects via the RFID systems (in Fig. 1). Our idea is based on the observation that by attaching multiple tags onto the surface of the packages, we are able to derive the 3D orientation of each single package and the 3D stacking situation of multiple packages according to the backscattered RF-signals from these reference tags. Our approach of RF-3DScan is as follows: We attach a set of reference tags on the surface of the packages, then we utilize a single RFID antenna to continuously scan the tagged packages, while the antenna is moving along

straight lines within a constrained 2-dimensional space.As the II.RELATED WORK antenna is moving,by extracting the phase differences from A.Computer Vision and Sensor-based Approach the specified tags at different time points,we build the angle Computer-vision-based solutions mainly leverage the depth profiles to depict the geometry angles between the antenna-tag camera to perform 3D reconstruction of multiple objects pairs.By comparing the angle profiles of different reference [1,2].To avoid the blind angles in 3D reconstruction for the tags,we are able to derive the relative positions of these tags specified objects.usually multiple depth cameras are deployed on the specified package,and further figure out the package at different positions to perform multi-view reconstruction for orientation and package stacking for multiple packages. their 3D models [1],or a moving depth camera is used to There are three key challenges to realize 3D reconstruction build the 3D models in a mobile approach [2].In a word, via RFID systems.The first challenge is to determine the these approaches suffer from the line-of-sight (LOS)constraint package orientation according to the RF-signals from the reference tags attached to a specified package.To tackle this in 3D perception,and they are vulnerable to the limitation of the light intensity.Sensor-based solutions [3,4]mainly challenge,we extract angle profiles from the phases of the attach the battery-powered sensors (such as inertial sensors or RF-signals,then we build an angle-profile-based model to GPS modules)to the surface of the objects,and continuously transform the RF-signals into the indicators for the relative monitor the 3D placement of the specified objects,so as to localization among the reference tags.Thus,after performing track the orientation variation [3],or the stacking situation 1-dimensional mobile scanning along a straight line,we can among multiple objects. determine the relative positions of the reference tag pairs on the package,and use this information to further derive the B.RFID-based Approach package orientation.The second challenge is to determine Orientation tracking:By attaching multiple RFID tags onto the stacking situation among multiple packages,according to the specified object,it is possible to track the orientation the RF-signals from the reference tags attached to multiple variation of the object according to the variation of the packages.To tackle this challenge,we further perform a 2- corresponding RF-signals [5,6].Tagball [5]is proposed as dimensional mobile scanning to scan the packages along the a 3D human-computer interaction system,where multiple orthogonal direction of the previous scanning direction,such passive tags are attached to a controlling ball,such that the that the relative 3D positions of the reference tags from motions of the ball rotation from users can be detected from different packages can be determined.In this way,we can the phase changes of multiple tags.Tagyro [6]attaches an estimate the centers of packages according to the reference array of passive RFID tags as orientation sensors on the tags,and derive the relative locations of different packages in objects,by transforming the runtime phase offsets between the 3D space.The third challenge is to select effective refer- tags into the orientation angle.Compared with our RF-3DScan ence tag pairs for accurately deriving the package orientation system,these approaches track the orientation variation of the and stacking situation.To tackle this challenge,we filter out dynamically moving objects,whereas our approach aims to those reference tags with unstable phases.which are located determine the orientation of statically placed packages. outside the field of major antenna beam during scanning,by Localization:RFID localization generally falls into two cat- referring to the received signal strength(RSS).Further,as our egories:absolute localization [7-10]and relative localization empirical study shows that the absolute phase of the RF-signal [11-15].By attaching multiple tags and pinpointing each tag's varies with different orientations of the reference tag,thus we 3D coordinates,the absolute localization can be tailored to our measure the phase differences to extract the angle profiles from problem for 3D reconstruction.However,this approach suffers the reference tags during the mobile scanning. from complicated system deployment and collaboration.For To the best of our knowledge,this paper presents the example,the state-of-the-art absolute localization schemes first study of using RFID for 3D reconstruction on tagged PinIt [7]and Tagoram [10]are able to achieve cm-level packages.We make three contributions as follows.1)For 3D localization accuracy,however,they either need to deploy reconstruction on the packages,we attach multiple reference many reference tags or require sophisticated calibration of RFID tags onto the packages,and respectively tackle the issues multiple readers.Rather than absolute localization,recent of package orientation and package stacking,by leveraging RFID researches start to focus on the relative localization the angle profiles extracted from the RF-signals.We build an of multiple objects without any pre-deployment of reference angle-profile-based model to depict the relationship between nodes.Relative localization investigates the relative locations the RF-signals from the reference tags and the package ori- of a set of objects as oppose to their absolute coordinates entation/stacking.2)We propose a mobile scanning solution STPP [13]is the first work to tackle 2D relative localization.It to realize the 3D reconstruction of tagged packages.We are investigates the spatial-temporal dynamics in the phase profiles able to determine the package orientation via /-dimensional However,this approach leverages large-range scanning to mobile scanning,and further determine the package stacking detect the Vzone from the phase sequences,as it requires the via 2-dimensional mobile scanning.3)We have implemented a antenna to cross the perpendicular point during the scanning prototype system to evaluate the performance,the experiment to collect enough phases.Compared with STPP,our approach results in real settings show that RF-3DScan achieves about performs 3D relative localization by leveraging the angle 92.5%bottom face accuracy and about 4.08 angle error. profiles from rather small-range scanning

straight lines within a constrained 2-dimensional space. As the antenna is moving, by extracting the phase differences from the specified tags at different time points, we build the angle profiles to depict the geometry angles between the antenna-tag pairs. By comparing the angle profiles of different reference tags, we are able to derive the relative positions of these tags on the specified package, and further figure out the package orientation and package stacking for multiple packages. There are three key challenges to realize 3D reconstruction via RFID systems. The first challenge is to determine the package orientation according to the RF-signals from the reference tags attached to a specified package. To tackle this challenge, we extract angle profiles from the phases of the RF-signals, then we build an angle-profile-based model to transform the RF-signals into the indicators for the relative localization among the reference tags. Thus, after performing 1-dimensional mobile scanning along a straight line, we can determine the relative positions of the reference tag pairs on the package, and use this information to further derive the package orientation. The second challenge is to determine the stacking situation among multiple packages, according to the RF-signals from the reference tags attached to multiple packages. To tackle this challenge, we further perform a 2- dimensional mobile scanning to scan the packages along the orthogonal direction of the previous scanning direction, such that the relative 3D positions of the reference tags from different packages can be determined. In this way, we can estimate the centers of packages according to the reference tags, and derive the relative locations of different packages in the 3D space. The third challenge is to select effective refer￾ence tag pairs for accurately deriving the package orientation and stacking situation. To tackle this challenge, we filter out those reference tags with unstable phases, which are located outside the field of major antenna beam during scanning, by referring to the received signal strength (RSS). Further, as our empirical study shows that the absolute phase of the RF-signal varies with different orientations of the reference tag, thus we measure the phase differences to extract the angle profiles from the reference tags during the mobile scanning. To the best of our knowledge, this paper presents the first study of using RFID for 3D reconstruction on tagged packages. We make three contributions as follows. 1) For 3D reconstruction on the packages, we attach multiple reference RFID tags onto the packages, and respectively tackle the issues of package orientation and package stacking, by leveraging the angle profiles extracted from the RF-signals. We build an angle-profile-based model to depict the relationship between the RF-signals from the reference tags and the package ori￾entation/stacking. 2) We propose a mobile scanning solution to realize the 3D reconstruction of tagged packages. We are able to determine the package orientation via 1-dimensional mobile scanning, and further determine the package stacking via 2-dimensional mobile scanning. 3) We have implemented a prototype system to evaluate the performance, the experiment results in real settings show that RF-3DScan achieves about 92.5% bottom face accuracy and about 4.08◦ angle error. II. RELATED WORK A. Computer Vision and Sensor-based Approach Computer-vision-based solutions mainly leverage the depth camera to perform 3D reconstruction of multiple objects [1, 2]. To avoid the blind angles in 3D reconstruction for the specified objects, usually multiple depth cameras are deployed at different positions to perform multi-view reconstruction for their 3D models [1], or a moving depth camera is used to build the 3D models in a mobile approach [2]. In a word, these approaches suffer from the line-of-sight (LOS) constraint in 3D perception, and they are vulnerable to the limitation of the light intensity. Sensor-based solutions [3, 4] mainly attach the battery-powered sensors (such as inertial sensors or GPS modules) to the surface of the objects, and continuously monitor the 3D placement of the specified objects, so as to track the orientation variation [3], or the stacking situation among multiple objects. B. RFID-based Approach Orientation tracking: By attaching multiple RFID tags onto the specified object, it is possible to track the orientation variation of the object according to the variation of the corresponding RF-signals [5, 6]. Tagball [5] is proposed as a 3D human-computer interaction system, where multiple passive tags are attached to a controlling ball, such that the motions of the ball rotation from users can be detected from the phase changes of multiple tags. Tagyro [6] attaches an array of passive RFID tags as orientation sensors on the objects, by transforming the runtime phase offsets between tags into the orientation angle. Compared with our RF-3DScan system, these approaches track the orientation variation of the dynamically moving objects, whereas our approach aims to determine the orientation of statically placed packages. Localization: RFID localization generally falls into two cat￾egories: absolute localization [7–10] and relative localization [11–15]. By attaching multiple tags and pinpointing each tag’s 3D coordinates, the absolute localization can be tailored to our problem for 3D reconstruction. However, this approach suffers from complicated system deployment and collaboration. For example, the state-of-the-art absolute localization schemes PinIt [7] and Tagoram [10] are able to achieve cm-level localization accuracy, however, they either need to deploy many reference tags or require sophisticated calibration of multiple readers. Rather than absolute localization, recent RFID researches start to focus on the relative localization of multiple objects without any pre-deployment of reference nodes. Relative localization investigates the relative locations of a set of objects as oppose to their absolute coordinates. STPP [13] is the first work to tackle 2D relative localization. It investigates the spatial-temporal dynamics in the phase profiles However, this approach leverages large-range scanning to detect the V-zone from the phase sequences, as it requires the antenna to cross the perpendicular point during the scanning to collect enough phases. Compared with STPP, our approach performs 3D relative localization by leveraging the angle profiles from rather small-range scanning

△d △d d x cosa d x cosa 0 60 120180240300360 Moving Rotation(deg.）】 direction (a)rotate along the Z axis (b)phase change during rotation A Pi d Fig.2.Measured phase of a single tag rotating along the Z axis Fig.3.AoA in static scanning Fig.4.AoA in mobile scanning III.ANGLE-PROFILE-BASED MODELING The phase difference is related to the distance difference A.Limitations of Phase-based Measurement from the tag to the antennas.When h>d,the relationship between the phase difference (A0 =0A:-0A2+n.n means The RF phase is a widely used attribute of the wireless the phase offset caused by the hardware characteristics of A signal,ranging from 0 to 2.Due to the ultra-high working and A2)and the distance difference (Ad dr.A-dr.A2 frequency (indicating short wave-length)in RFIDs and fine- grained measure resolution of phase value by COTS readers, d cos a)can be approximated as: the phase is very sensitive to the tag-antenna distance,which 2 d cos a△6 2π +n (2) gives us the potential chance to achieve accurate 3D recon- 入 struction.Suppose dis is the distance between the antenna wherencan be any integer in[-共-2，共-]，its range and the tag.Since the backscatter communication of RFID is 4.When d<the value range of n is smalier than 1, is round-trip,the signal totally traverses a distance of 2dis which means n has a unique value,so a is deterministic. in each communication.Besides the distance,some hardware 2)Angle in Mobile Scanning:With respect to multiple characteristics will also distort the phase value.Hence,the antennas,the phase offsets related to their own hardware phase 6 reported by the reader can be expressed as: characteristics are different,so it is hard to determine n Hence,we prefer a mobile antenna to multiple static antennas, 2元 ×2dis+n mod 2 (1) in which case the can be canceled. For a mobile antenna,the angle-of-arrival is a little different. where A is the wavelength,n represents the phase offset caused Without the loss of generality,we redefine the AoA in a mobile by the hardware characteristics.Although the phase accurately case,as shown in Fig.4.Similarly,T is the tag position reflects the distance,we face three challenges before putting and V is its projected point on the antenna moving line, into use:1)The distort factor n is unknown;2)The phase value its perpendicular distance is h.Let the mobile antenna be repeats periodically,it is not feasible to use it directly;3)In at position A,then the included angle of line TA and the addition to dis and n,our extensive experiments show that the antenna moving direction is just the angle-of-arrival (a)for tag orientation influences the phase value 0.Fig.2 plots the the tag when the antenna is at position A. phase change as a tag rotates along the Z axis,as the phase To estimate the angle at position A,we only need the phases varies continuously over the rotation.Next,we discuss how collected at the two nearby positions (P and P2),centered to use the angle-of-arrival approach to overcome above three on the antenna (PA =AP2).Thus,the phase difference at challenges,and benefit our system design in the sequel. position P and P2 can be used to estimate a with Eg.2.By combining the angles at different antenna positions,we can B.Angle Profile derive an angle profile for a specified tag. Angle-of-Arrival (AoA)is one of the most popular RF- C.Metrics of Angle Profile based localization measurements using phase difference.The Suppose there are two tags and one antenna in the same basic idea of our approach is that by moving the antenna plane(Fig.5).The antenna moves linearly from O to A,so to scan the tags,we extract the phase differences from the it passes through T first,followed by T2.When the antenna specified tags at different time points,then we derive the passes through the tag (corresponding to point V in Fig.4), geometry angles between the tag-antenna pairs at different the angle-of-arrival (o)of that tag reaches /2,naming this positions,which is called angle profile. point as the perpendicular point.Similarly,we call the distance 1)Angle in Static Scanning:As shown in Fig.3,a tag is from the tag to the perpendicular point perpendicular distance, set at T,A1 and A2 are two antennas separated by d,M is the direction perpendicular to the antenna moving direction as the middle point of A1A2.V is the projected point of T on perpendicular direction.As Ti is on the left along the antenna the tag pair line A1A2,the perpendicular distance is h.The moving direction,its perpendicular point shows earlier than included angle between line TM and line MV is the AoA T2.Hence,the perpendicular point is the key metric for the for tag T,denoted as a.Let dr.A,and dr.A represent the tags'relative positions along the moving direction. distances between T and the antennas,the antennas collect the Besides the perpendicular point,there is the other special phases as 6A and 042 respectively.641,6A2 E0,27). point:equal angle point.The equal angle point is where the

   (a) rotate along the Z axis        5RWDWLRQGHJ         3KDVHUDG (b) phase change during rotation Fig. 2. Measured phase of a single tag rotating along the Z axis ܶ ߙ ο݀ ൎ ߙ•‘… ൈ݀ ଶܣ ଵܣ ݄ ܸ ܯ ݀ Fig. 3. AoA in static scanning ܣ ܶ ߙ ο݀ ൎ ߙ•‘… ൈ݀ ܲଵ ܲଶ ݄ ܸ ݀ ݃݊݅ݒ݋ܯ †‹‡…‹‘ Fig. 4. AoA in mobile scanning III. ANGLE-PROFILE-BASED MODELING A. Limitations of Phase-based Measurement The RF phase is a widely used attribute of the wireless signal , ranging from 0 to 2π. Due to the ultra-high working frequency (indicating short wave-length) in RFIDs and fine￾grained measure resolution of phase value by COTS readers, the phase is very sensitive to the tag-antenna distance, which gives us the potential chance to achieve accurate 3D recon￾struction. Suppose dis is the distance between the antenna and the tag. Since the backscatter communication of RFID is round-trip, the signal totally traverses a distance of 2dis in each communication. Besides the distance, some hardware characteristics will also distort the phase value. Hence, the phase θ reported by the reader can be expressed as: θ = 2π λ × 2dis + η mod 2π (1) where λ is the wavelength, η represents the phase offset caused by the hardware characteristics. Although the phase accurately reflects the distance, we face three challenges before putting into use: 1) The distort factor η is unknown; 2) The phase value repeats periodically, it is not feasible to use it directly; 3) In addition to dis and η, our extensive experiments show that the tag orientation influences the phase value θ. Fig. 2 plots the phase change as a tag rotates along the Z axis, as the phase varies continuously over the rotation. Next, we discuss how to use the angle-of-arrival approach to overcome above three challenges, and benefit our system design in the sequel. B. Angle Profile Angle-of-Arrival (AoA) is one of the most popular RF￾based localization measurements using phase difference. The basic idea of our approach is that by moving the antenna to scan the tags, we extract the phase differences from the specified tags at different time points, then we derive the geometry angles between the tag-antenna pairs at different positions, which is called angle profile. 1) Angle in Static Scanning: As shown in Fig. 3, a tag is set at T, A1 and A2 are two antennas separated by d, M is the middle point of A1A2. V is the projected point of T on the tag pair line A1A2, the perpendicular distance is h. The included angle between line TM and line MV is the AoA for tag T, denoted as α. Let dT ,A1 and dT ,A2 represent the distances between T and the antennas, the antennas collect the phases as θA1 and θA2 respectively. θA1 , θA2 ∈ [0, 2π). The phase difference is related to the distance difference from the tag to the antennas. When h d, the relationship between the phase difference (Δθ = θA1 −θA2 +θη, θη means the phase offset caused by the hardware characteristics of A1 and A2) and the distance difference (Δd = dT ,A1 − dT ,A2  d cos α) can be approximated as: 2d cos α λ = Δθ 2π + n (2) where n can be any integer in  −2d λ − Δθ 2π , 2d λ − Δθ 2π  , its range is 4d λ . When d < λ 4 , the value range of n is smaller than 1, which means n has a unique value, so α is deterministic. 2) Angle in Mobile Scanning: With respect to multiple antennas, the phase offsets related to their own hardware characteristics are different, so it is hard to determine θη. Hence, we prefer a mobile antenna to multiple static antennas, in which case the θη can be canceled. For a mobile antenna, the angle-of-arrival is a little different. Without the loss of generality, we redefine the AoA in a mobile case, as shown in Fig. 4. Similarly, T is the tag position and V is its projected point on the antenna moving line, its perpendicular distance is h. Let the mobile antenna be at position A, then the included angle of line T A and the antenna moving direction is just the angle-of-arrival (α) for the tag when the antenna is at position A. To estimate the angle at position A, we only need the phases collected at the two nearby positions (P1 and P2), centered on the antenna (P1A = AP2). Thus, the phase difference at position P1 and P2 can be used to estimate α with Eq. 2. By combining the angles at different antenna positions, we can derive an angle profile for a specified tag. C. Metrics of Angle Profile Suppose there are two tags and one antenna in the same plane (Fig. 5). The antenna moves linearly from O to A, so it passes through T1 first, followed by T2. When the antenna passes through the tag (corresponding to point V in Fig. 4), the angle-of-arrival (α) of that tag reaches π/2, naming this point as the perpendicular point. Similarly, we call the distance from the tag to the perpendicular point perpendicular distance, the direction perpendicular to the antenna moving direction as perpendicular direction. As T1 is on the left along the antenna moving direction, its perpendicular point shows earlier than T2. Hence, the perpendicular point is the key metric for the tags’ relative positions along the moving direction. Besides the perpendicular point, there is the other special point: equal angle point. The equal angle point is where the

Perpendicular distance Preprocess Determine Determine package orientation package stacking RF Angle compuation (single) (multiple) Signals Angle smoothing T ·Bottom/top face Relative stacking Lincar fitin Rotation angle situation 2 Moving Fig.7. Architecture of RF-3DScan 0 Equal angle Perpendicular A distance between two zero points just reflects the tags'perpendicular point point points separation.In addition,the intersection of the two lines Fig.5. Metrics for the angle profiles represents the position where the tags are projected on the cota same line with the antenna,corresponding to the equal angle point.Specially,the smaller h is,the larger is,and the Equal angle point sharper the line is.As the h of Ti is smaller than T2,thelkll T2 of Ti is larger,so the line of Ti decreases faster than T2. Perpendicular point For a certain tag,its angle profile records its angles at T Moving different positions.Note that,for analyzing the change of distance the angles,it is the separation between the positions and Fig.6.Model of the angle profiles the corresponding change of cot o that matter,so the moved antenna and the two tags are in the same line,so n and 72 distance l does not necessarily the actual moved distance of the share the same angle.Before equal point,the angle of Ti is antenna.That is,we have no constrict to the coordinates of the smaller than the angle of T2.On the contrary,the angle of T positions,as long as they refer to the same basis.For example, changes to be bigger than that of T2 after the equal angle point. let the antenna moves at a constant speed,and set a random No matter for T or T2,its angle increases continuously during time as the starting moving time.When the antenna collects the antenna moving process,so it is obvious that the angle of phase during the moving process,it records the corresponding time as well,then that position can be estimated with the Ti changes faster than that of T2.Such phenomenon is due to the smaller perpendicular distance of T1.Thus,according time interval and the constant moving speed.So,in Eq.4, to the angle change rate,we can determine the tags'relative the angle a and the moved distance I are known parameters, positions along the perpendicular direction. there're two remaining unknown parameters:h and lo,which can be estimated by linear fitting with multiple angles during D.Model of Angle Profile the moving process in the angle profile.Meanwhile,lo depends To depict the angle-profile-based measurement metrics in on when the antenna passes through the tag.For different mathematics,we build a linear model to derive the metrics tags,the antenna start point should be the same one (as they from the angle profile automatically.Considering Fig.4,the share the same starting time),so the larger lo is.the later that angle-of-arrival can be expressed as: line reaches 0,and the tag is more ahead along the antenna cot a=y-yA moving direction.Thus,by leveraging these properties,we can (3) h determine the tags'relative positions,as: where cot means the cotangent function,h is the perpendicular 1)The value of reflects the perpendicular distance from distance between the tag and the antenna moving trace.yA the tag to the antenna moving trace:the larger is, and yv represent the coordinates of point A and V along the the smaller the perpendicular distance is. antenna moving direction.Assume there is an antenna start 2)The value of lo determines the projected position of the point S,the distance from S to V is lo,the antenna moved corresponding tag along the antenna moving direction. distance be 1.Thus,(lo-l)represents the distance from the The difference of lo between two tags indicates their antenna to the perpendicular point (same as (yv-yA)),the interval in the antenna moving direction. angle can be rewritten as: IV.SYSTEM OVERVIEW cot a=kl +b,k = h6= lo (4) RF-3DScan is a 3D reconstruction system for tagged pack- h ages via RFID technology.For RF-3DScan,the geometry where the scope k is related to the minus reciprocal of h,the relationships of the tags attached on a single package is known intercept b depends on the ratio of lo and h. as priori and the tag deployment obeys two important rules. Taking the tags in Fig.5,the transformed angle expression Also,we make the following assumptions:1)The antenna using Eq.4 should look like the lines shown in Fig.6.As I moves at a constant speed;2)Each package is a standard cube. increases continuously during the moving process,o increases and they are fully on the ground or parallel to the ground (on as well.When the antenna reaches the perpendicular point, the ground is the special case of parallel to the ground). a is equal to /2,so cota =0.The line of T reaches Fig.7 illustrates the architecture of RF-3DScan.RF-3DScan 0 earlier than T2.Thus,the order of such zero points are takes the RF-signals from the tags as input,then outputs 3D corresponding to the tags perpendicular points,and the spacing profiles for multiple packages.The whole system consists

ଵߙ ଶߙ ଵԢߙ ଶԢߙ ‡’‡†‹…Žƒ †‹•ƒ…‡ ‘‹‰ †‹•ƒ…‡ ݄ଶ ƒŽƒ‰Ž‡ ’‘‹ ‡’‡†‹…Žƒ ’‘‹ ݄ଵ ܶଶ ܶଵ ܣ ܱ Fig. 5. Metrics for the angle profiles ܶଶ ܶଵ ߙ ‘… ƒŽƒ‰Ž‡’‘‹ ‘‹‰ †‹•ƒ…‡ Ͳ ‡’‡†‹…Žƒ’‘‹ Fig. 6. Model of the angle profiles antenna and the two tags are in the same line, so T1 and T2 share the same angle. Before equal point, the angle of T1 is smaller than the angle of T2. On the contrary, the angle of T1 changes to be bigger than that of T2 after the equal angle point. No matter for T1 or T2, its angle increases continuously during the antenna moving process, so it is obvious that the angle of T1 changes faster than that of T2. Such phenomenon is due to the smaller perpendicular distance of T1. Thus, according to the angle change rate, we can determine the tags’ relative positions along the perpendicular direction. D. Model of Angle Profile To depict the angle-profile-based measurement metrics in mathematics, we build a linear model to derive the metrics from the angle profile automatically. Considering Fig. 4, the angle-of-arrival can be expressed as: cot α = yV − yA h (3) where cot means the cotangent function, h is the perpendicular distance between the tag and the antenna moving trace. yA and yV represent the coordinates of point A and V along the antenna moving direction. Assume there is an antenna start point S, the distance from S to V is l0, the antenna moved distance be l. Thus, (l0 − l) represents the distance from the antenna to the perpendicular point (same as (yV − yA)), the angle can be rewritten as: cot α = kl + b, k = − 1 h, b = l0 h (4) where the scope k is related to the minus reciprocal of h, the intercept b depends on the ratio of l0 and h. Taking the tags in Fig. 5, the transformed angle expression using Eq. 4 should look like the lines shown in Fig. 6. As l increases continuously during the moving process, α increases as well. When the antenna reaches the perpendicular point, α is equal to π/2, so cot α = 0. The line of T1 reaches 0 earlier than T2. Thus, the order of such zero points are corresponding to the tags perpendicular points, and the spacing 3UHSURFHVV $QJOHFRPSXWDWLRQ /LQHDUILWWLQJ $QJOHVPRRWKLQJ 5) 6LJQDOV 'HWHUPLQH SDFNDJHRULHQWDWLRQ VLQJOH 'HWHUPLQH SDFNDJHVWDFNLQJ PXOWLSOH ‡ %RWWRPWRSIDFH ‡ 5RWDWLRQDQJOH 5HODWLYHVWDFNLQJ VLWXDWLRQ Fig. 7. Architecture of RF-3DScan between two zero points just reflects the tags’ perpendicular points separation. In addition, the intersection of the two lines represents the position where the tags are projected on the same line with the antenna, corresponding to the equal angle point. Specially, the smaller h is, the larger k is, and the sharper the line is. As the h of T1 is smaller than T2, the k of T1 is larger, so the line of T1 decreases faster than T2. For a certain tag, its angle profile records its angles at different positions. Note that, for analyzing the change of the angles, it is the separation between the positions and the corresponding change of cot α that matter, so the moved distance l does not necessarily the actual moved distance of the antenna. That is, we have no constrict to the coordinates of the positions, as long as they refer to the same basis. For example, let the antenna moves at a constant speed, and set a random time as the starting moving time. When the antenna collects phase during the moving process, it records the corresponding time as well, then that position can be estimated with the time interval and the constant moving speed. So, in Eq. 4, the angle α and the moved distance l are known parameters, there’re two remaining unknown parameters: h and l0, which can be estimated by linear fitting with multiple angles during the moving process in the angle profile. Meanwhile, l0 depends on when the antenna passes through the tag. For different tags, the antenna start point should be the same one (as they share the same starting time), so the larger l0 is, the later that line reaches 0, and the tag is more ahead along the antenna moving direction. Thus, by leveraging these properties, we can determine the tags’ relative positions, as: 1) The value of k reflects the perpendicular distance from the tag to the antenna moving trace: the larger k is, the smaller the perpendicular distance is. 2) The value of l0 determines the projected position of the corresponding tag along the antenna moving direction. The difference of l0 between two tags indicates their interval in the antenna moving direction. IV. SYSTEM OVERVIEW RF-3DScan is a 3D reconstruction system for tagged pack￾ages via RFID technology. For RF-3DScan, the geometry relationships of the tags attached on a single package is known as priori and the tag deployment obeys two important rules. Also, we make the following assumptions: 1) The antenna moves at a constant speed; 2) Each package is a standard cube, and they are fully on the ground or parallel to the ground (on the ground is the special case of parallel to the ground). Fig. 7 illustrates the architecture of RF-3DScan. RF-3DScan takes the RF-signals from the tags as input, then outputs 3D profiles for multiple packages. The whole system consists

of three components:1)Preprocess:with the RF-signals of VI.DETERMINE PACKAGE ORIENTATION the tags,RF-3DScan builds the angle profiles by using the FOR EACH SINGLE PACKAGE phase differences at different time points for each tag,and To reconstruct a single package,it makes the same sense extracts the indicators in the angle profiles for the relative to determine the package orientation.so we just need to localization among the tags by linear fitting.2)Determine identify the bottom face of this package,and estimate the package orientation for a single package:by comparing the relative rotation angle of the vertical sides against the antenna relative positions of the tags on a specified package,RF- plane in a specified coordinate system.The basic idea of our 3DScan can determine which side of the package is on the approach is that we attach a set of passive RFID tags on the ground,and then evaluates the angle of the vertical sides in package under special rules,then employ one antenna to do a specified coordinate system.3)Determine package stacking 1-dimensional mobile scanning to build the angle profiles for for multiple packages:after deriving the orientation of a single each tag.Next,we compare them to determine the relative package,the centers of the packages are also determined positions of the tags,thus we can further realize the 3D along the scanning direction.By performing a 2-dimensional reconstruction for a single package. mobile scanning,RF-3DScan combines the results from the two orthogonal scanning,so the relative positions of these A.Deploy Reference Tags packages in the 3D space can be determined. In order to determine the package orientation only by 1- dimensional mobile scanning,we need to deploy the tags in an V.DATA PREPROCESS efficient way.The design principle of the tag deployment is to With the raw RF-signals,we need to build angle profiles of use as fewer as tags to depict the package uniquely,accurately different tags first.The preprocessing can be divided into three and conveniently.So,we make two rules as follows: steps:angle computation,angle smoothing and linear fitting. Rule 1:The orientations of the tags should be along different A.Angle Computation orthogonal axes.As the package can be with any orientation As the antenna collects phases at different time points in the 3D space,we should pay attention to ensuring there are always enough effective tags reflecting the signals to the during its mobile scanning,we can extract the phase dif- ferences for a tag at different positions.Using these phase antenna.As the 3D space can be defined with three orthogonal differences,we compute the angle-of-arrivals with Eg.2.To axes,we can just let the tags deployed along these axes alone, get a deterministic angle,as mentioned above,the separation as shown in Fig.8(a).With this rule,tags along one direction of the positions among two phases should be within A/4. at most are in the blind direction,so other tags can get enough power to reflect their signals to the antenna effectively. B.Angle Smoothing Rule 2:The tags should be deployed along different orthog- Although using the phase difference from the two positions onal axes.As for identifying the bottom face of the package with the small separation can get a unique angle,the noise like it is the same to find which tags are along the vertical axis and multi-path effect would influence the measured phases,there what order these tags are.So,there should be at least three would exist large fluctuation in angles,so angle smoothing is tag pairs (four tags)along three orthogonal axes separately,as required.Usually,the phases collected by the antenna is not shown in Fig.8(b).Under this rule,whatever the orientation uniform,so is the angle distribution,thus it is not suitable of the package is,there is always one tag pair along the to use the common smooth algorithms,e.g.,low-pass filter. △9+y十 vertical axis,so we can transform the identification of the Taking the noise u into consideration:cosa =2d2 bottom face of the package into finding the vertical tag pair when d is very small,has much influence on cosa. and determining their orders along the vertical axis. While when d increases,such distortion effect decreases,but Combining the above two rules,Fig.8(c)illustrates a there exist redundant angles in the results,only one of them possible tag deployment satisfying these two rules.No matter is the true value.Hence,we can derive two sets of angles what orientation the package is,there are four tags at least to from two phase separations:a small one and a large one,then avoid the signal blind direction.Also,there is always one tag use the unique angles from the small separation to filter the pair along the Z axis.By determining the tags'order of this several angle candidates from the large separation,thus,we get tag pair,we can derive which side of the package is on the a relative accurate angle profile with less fluctuation [12].Note ground then. that,too large separation will bring too much environmental B.Determine Package Orientation change and break the restraint of the angle estimation method. Thus,we set the small separation around 5-8cm and the large To determine the package orientation,it demands to identify the bottom face of the package and the relative angle of the separation within 15cm empirically when the antenna is in front of the packages about 1m. vertical sides in a specified coordinate system.Considering our assumption that one side of the cube package must be C.Linear Fitting parallel to the ground,when we deploy the tags of a package With the smoothed angles at different positions from an like the solution described in Fig.8(c),we can identify which angle profile for a certain tag,we can use the linear model as tag pair is along the Z axis and what order the tag pair is Eq.4 to fit them,then derive the two important indicators(h instead.In this case,let the antenna do mobile scanning along and lo)of that tag for the later relative localization comparison. the X or Y axis only once,we can determine the orientation

of three components: 1) Preprocess: with the RF-signals of the tags, RF-3DScan builds the angle profiles by using the phase differences at different time points for each tag, and extracts the indicators in the angle profiles for the relative localization among the tags by linear fitting. 2) Determine package orientation for a single package: by comparing the relative positions of the tags on a specified package, RF- 3DScan can determine which side of the package is on the ground, and then evaluates the angle of the vertical sides in a specified coordinate system. 3) Determine package stacking for multiple packages: after deriving the orientation of a single package, the centers of the packages are also determined along the scanning direction. By performing a 2-dimensional mobile scanning, RF-3DScan combines the results from the two orthogonal scanning, so the relative positions of these packages in the 3D space can be determined. V. DATA PREPROCESS With the raw RF-signals, we need to build angle profiles of different tags first. The preprocessing can be divided into three steps: angle computation, angle smoothing and linear fitting. A. Angle Computation As the antenna collects phases at different time points during its mobile scanning, we can extract the phase dif￾ferences for a tag at different positions. Using these phase differences, we compute the angle-of-arrivals with Eq. 2. To get a deterministic angle, as mentioned above, the separation of the positions among two phases should be within λ/4. B. Angle Smoothing Although using the phase difference from the two positions with the small separation can get a unique angle, the noise like multi-path effect would influence the measured phases, there would exist large fluctuation in angles, so angle smoothing is required. Usually, the phases collected by the antenna is not uniform, so is the angle distribution, thus it is not suitable to use the common smooth algorithms, e.g., low-pass filter. Taking the noise μ into consideration: cos α = λ 2d Δθ+μ 2π + nλ 2d , when d is very small, μ has much influence on cos α. While when d increases, such distortion effect decreases, but there exist redundant angles in the results, only one of them is the true value. Hence, we can derive two sets of angles from two phase separations: a small one and a large one, then use the unique angles from the small separation to filter the several angle candidates from the large separation, thus, we get a relative accurate angle profile with less fluctuation [12]. Note that, too large separation will bring too much environmental change and break the restraint of the angle estimation method. Thus, we set the small separation around 5-8cm and the large separation within 15cm empirically when the antenna is in front of the packages about 1m. C. Linear Fitting With the smoothed angles at different positions from an angle profile for a certain tag, we can use the linear model as Eq. 4 to fit them, then derive the two important indicators (h and l0) of that tag for the later relative localization comparison. VI. DETERMINE PACKAGE ORIENTATION FOR EACH SINGLE PACKAGE To reconstruct a single package, it makes the same sense to determine the package orientation, so we just need to identify the bottom face of this package, and estimate the relative rotation angle of the vertical sides against the antenna plane in a specified coordinate system. The basic idea of our approach is that we attach a set of passive RFID tags on the package under special rules, then employ one antenna to do 1-dimensional mobile scanning to build the angle profiles for each tag. Next, we compare them to determine the relative positions of the tags, thus we can further realize the 3D reconstruction for a single package. A. Deploy Reference Tags In order to determine the package orientation only by 1- dimensional mobile scanning, we need to deploy the tags in an efficient way. The design principle of the tag deployment is to use as fewer as tags to depict the package uniquely, accurately and conveniently. So, we make two rules as follows: Rule 1: The orientations of the tags should be along different orthogonal axes. As the package can be with any orientation in the 3D space, we should pay attention to ensuring there are always enough effective tags reflecting the signals to the antenna. As the 3D space can be defined with three orthogonal axes, we can just let the tags deployed along these axes alone, as shown in Fig. 8(a). With this rule, tags along one direction at most are in the blind direction, so other tags can get enough power to reflect their signals to the antenna effectively. Rule 2: The tags should be deployed along different orthog￾onal axes. As for identifying the bottom face of the package, it is the same to find which tags are along the vertical axis and what order these tags are. So, there should be at least three tag pairs (four tags) along three orthogonal axes separately, as shown in Fig. 8(b). Under this rule, whatever the orientation of the package is, there is always one tag pair along the vertical axis, so we can transform the identification of the bottom face of the package into finding the vertical tag pair and determining their orders along the vertical axis. Combining the above two rules, Fig. 8(c) illustrates a possible tag deployment satisfying these two rules. No matter what orientation the package is, there are four tags at least to avoid the signal blind direction. Also, there is always one tag pair along the Z axis. By determining the tags’ order of this tag pair, we can derive which side of the package is on the ground then. B. Determine Package Orientation To determine the package orientation, it demands to identify the bottom face of the package and the relative angle of the vertical sides in a specified coordinate system. Considering our assumption that one side of the cube package must be parallel to the ground, when we deploy the tags of a package like the solution described in Fig. 8(c), we can identify which tag pair is along the Z axis and what order the tag pair is instead. In this case, let the antenna do mobile scanning along the X or Y axis only once, we can determine the orientation

(a)rulel (b)rule2 (c)possible solution (a)possible casel (b)possible case2 (c)possible case3 Fig.8.Deploying reference tags Fig.9.Possible cases for the special package orientation for a single package.Note that,the antenna should be above or possible cases of the package orientation,as shown in Fig.8(c) below all the tags,and the package is at the same side against and Fig.9.Any of these possible cases can transform into the antenna plane during the scanning process. another case by rotating along the Y axis,but the relationships For simplicity,suppose the antenna is above all the tags of their perpendicular distances differ,so we can use these and ahead of all the tags along the X axis positive direction, relationships to vote for which case is the most possible case it moves along the Y axis,then the tag pair along the Z axis As we assume that the antenna is above all the tags and ahead is perpendicular to the antenna scanning direction.so their of all the tags along the X axis positive direction,let the perpendicular points should be the same.That is,the spacing perpendicular distances for T,T3,Ts be h,h3,hs,then for of their perpendicular points equals 0 in theory.Hence,if the case as Fig.8(c):h1h5,for the possible the spacing of the perpendicular points for a tag pair along casel (Fig.9(a)):hh3,h3 >h5.There are multiple tag pairs is a tag pair along the X axis,then there are two tag pairs for the comparison,here we list part of them for explanation. that their perpendicular points spacings are equal to 0,we will Then,by comparing the relationships of different tag pairs, discuss it later.After identifying which tag pair is along the we vote for the possible cases,and select the case with the Z axis,we can determine the tags'order by comparing their highest score as our estimation result. perpendicular distances extracted from their angle profiles.As C.Discussion the antenna is above all the tags,the tag with the smaller There must be a side of the package parallel to the ground: perpendicular distance of the vertical tag pair should be above As we assume that there must be a side of the packages the other along the Z axis.However,such comparison ignores parallel to the ground (which means the package is on the the relationships of the perpendicular distances for other tags, ground or on other packages,not leans).thus the state of the it is easy to make a wrong decision with only one comparison package is limited,the angle estimation is restricted to along results.Note that,the spacings of the perpendicular points of the Z axis.If not,the searching space of finding the optimal the tags should stay the same when the package is upside angle expands,the simple solution is to add one more mobile down,as the package rotates around the Y axis by 180.So scanning along the direction different from the previous one. we can estimate the relative angle of the package first,then the 3D reconstruction can be realized as well. use the relationships of the perpendicular distances among There may exist serious tag missing when many packages different tag pairs to vote for the tags'order of the vertical are stacked closely:As packages are stacked in storage,the tag pair,then determine the bottom face of the package.When large amounts of tags and small separations between the tags selecting the tag pairs among all the tags on the package,it is from different packages may cause the coupling effect or the significant to avoid the tags in the blind direction by filtering interrogation failure [16-20].But it exceeds the research fields the tags with relative weak RSS compared with other tags on of this paper,so we ignore it now. the package.The angle estimation is based on the spacings of The difference of the perpendicular distances for the tag the perpendicular points for different tag pairs,as: pair parallel to the antenna plane may be much smaller than that for the tag pair perpendicular to the antenna plane with arg min ∑d-6,(o)训 (5) the same tag spacing:As shown in Fig.10,the tag pair i=1 (T1,T2}is perpendicular to the antenna plane,while the where N is the number of tag pairs,is the spacing between tag pair (T2,T3}is parallel to the antenna plane.Their tag the perpendicular points of a tag pair by measurement,i() spacings are the same,as Adh=Ad,but their perpendicular represents the spacing at relative angle o theoretically distance differences are not similar.Suppose the distance from Now,considering the case shown in Fig.8(c)for example, Ti to the antenna is Im,the tag spacing is 0.2m,so the we illustrate how to deal with the special cases where there perpendicular distance difference between T2 and T3 is only are two tag pairs whose perpendicular point spacings are 1.65cm,which is much smaller than that between Ti and both equal to 0.As the antenna moves along the y axis, T2(20cm).Since the distance difference is so small,the the perpendicular points of the tag pair on the same surface relationship of such tag pair is probably to be wrong.So, [T3,Ta}or [T5,T6}are at the same point.With the relative it is better to set weights for the tag pairs based on their order of the tag pair [T1,T2}.as the tag Ti is on the left perpendicular distance differences when voting for multiple of T2 along the antenna moving direction,there are four possible package orientations

   (a) rule1    (b) rule2    ܶଵ ܶଶ ܶହ ܶ଺ ܶଷ ܶସ (c) possible solution Fig. 8. Deploying reference tags for a single package. Note that, the antenna should be above or below all the tags, and the package is at the same side against the antenna plane during the scanning process. For simplicity, suppose the antenna is above all the tags and ahead of all the tags along the X axis positive direction, it moves along the Y axis, then the tag pair along the Z axis is perpendicular to the antenna scanning direction, so their perpendicular points should be the same. That is, the spacing of their perpendicular points equals 0 in theory. Hence, if the spacing of the perpendicular points for a tag pair along a certain axis is 0, it is probable that the tag pair is along the Z axis now, except for some special cases where there is a tag pair along the X axis, then there are two tag pairs that their perpendicular points spacings are equal to 0, we will discuss it later. After identifying which tag pair is along the Z axis, we can determine the tags’ order by comparing their perpendicular distances extracted from their angle profiles. As the antenna is above all the tags, the tag with the smaller perpendicular distance of the vertical tag pair should be above the other along the Z axis. However, such comparison ignores the relationships of the perpendicular distances for other tags, it is easy to make a wrong decision with only one comparison results. Note that, the spacings of the perpendicular points of the tags should stay the same when the package is upside down, as the package rotates around the Y axis by 180◦. So we can estimate the relative angle of the package first, then use the relationships of the perpendicular distances among different tag pairs to vote for the tags’ order of the vertical tag pair, then determine the bottom face of the package. When selecting the tag pairs among all the tags on the package, it is significant to avoid the tags in the blind direction by filtering the tags with relative weak RSS compared with other tags on the package. The angle estimation is based on the spacings of the perpendicular points for different tag pairs, as: arg min φ  N i=1 δ i − δi(φ) (5) where N is the number of tag pairs, δ i is the spacing between the perpendicular points of a tag pair by measurement, δi(φ) represents the spacing at relative angle φ theoretically. Now, considering the case shown in Fig. 8(c) for example, we illustrate how to deal with the special cases where there are two tag pairs whose perpendicular point spacings are both equal to 0. As the antenna moves along the Y axis, the perpendicular points of the tag pair on the same surface {T3, T4} or {T5, T6} are at the same point. With the relative order of the tag pair {T1, T2}, as the tag T1 is on the left of T2 along the antenna moving direction, there are four ܶହ ܶ଺    ܶଵ ܶଶ ܶଷ ܶସ (a) possible case1 ܶଵ ܶଶ ܶ଺ ܶହ ܶଵ ܶ଺ ܶହ ܶଶ    ܶସ ܶଷ (b) possible case2 ܶଵ ܶଶ    ܶସ ܶଷ ܶ଺ ܶହ (c) possible case3 Fig. 9. Possible cases for the special package orientation possible cases of the package orientation, as shown in Fig. 8(c) and Fig. 9. Any of these possible cases can transform into another case by rotating along the Y axis, but the relationships of their perpendicular distances differ, so we can use these relationships to vote for which case is the most possible case. As we assume that the antenna is above all the tags and ahead of all the tags along the X axis positive direction, let the perpendicular distances for T1, T3, T5 be h1, h3, h5, then for the case as Fig. 8(c): h1 h5, for the possible case1 (Fig. 9(a)): h1 h3, h3 h3, h3 > h5. There are multiple tag pairs for the comparison, here we list part of them for explanation. Then, by comparing the relationships of different tag pairs, we vote for the possible cases, and select the case with the highest score as our estimation result. C. Discussion There must be a side of the package parallel to the ground: As we assume that there must be a side of the packages parallel to the ground (which means the package is on the ground or on other packages, not leans), thus the state of the package is limited, the angle estimation is restricted to along the Z axis. If not, the searching space of finding the optimal angle expands, the simple solution is to add one more mobile scanning along the direction different from the previous one, the 3D reconstruction can be realized as well. There may exist serious tag missing when many packages are stacked closely: As packages are stacked in storage, the large amounts of tags and small separations between the tags from different packages may cause the coupling effect or the interrogation failure [16–20]. But it exceeds the research fields of this paper, so we ignore it now. The difference of the perpendicular distances for the tag pair parallel to the antenna plane may be much smaller than that for the tag pair perpendicular to the antenna plane with the same tag spacing: As shown in Fig. 10, the tag pair {T1, T2} is perpendicular to the antenna plane, while the tag pair {T2, T3} is parallel to the antenna plane. Their tag spacings are the same, as Δdh = Δdv, but their perpendicular distance differences are not similar. Suppose the distance from T1 to the antenna is 1m, the tag spacing is 0.2m, so the perpendicular distance difference between T2 and T3 is only 1.65cm, which is much smaller than that between T1 and T2 (20cm). Since the distance difference is so small, the relationship of such tag pair is probably to be wrong. So, it is better to set weights for the tag pairs based on their perpendicular distance differences when voting for multiple possible package orientations.

Moving direction y Fig.10.The distance differences are different for Fig.11.Package stacking Fig.12.Determining package stacking in the 3D space the equal tag spacing along different directions VII.DETERMINE PACKAGE STACKING ↑Z ↑2 FOR MULTIPLE PACKAGES PP A.Limitations of the 1-dimensional Mobile Scanning for Ps Pa P7 P8 3 P Determining Package Stacking When we determine the package orientation,we derive y the indicators for the relative positions among the tags on a single package.As the geometry relationships of these tags are (a)along the X axis (b)along the Y axis known,we can combine these indicators (perpendicular dis- tance and perpendicular point)from different tags to estimate the indicators of the package's center point.Then.similarly. we compare the indicators of the center points of different P6 P7 Ps Pa packages to determine their stacking situation.Note that.when we determine the package orientation,we only need to perform 1-dimensional scanning,but it may not support the package (c)split plane stacking determination due to the 3-DoF in the 3D space.As (d)split space Fig.13.Illustration of determining the packages'order in the 3D space shown in Fig.11,suppose the antenna is above all the packages and is in front of the packages along the X axis,it moves along Combining their relative positions along the two directions. the y axis.It is easy to determine the packages'orders along the 3D space is divided into many pieces.For each piece,two the Y axis by referring to their perpendicular points,but it dimensions are fixed,the packages in it are lining up,so we may be a problem to determine their orders in the XZ plane. can use the perpendicular distances of the packages'centers If the packages line up,that is,the packages are along the to determine their orders in the piece.Thus,all the packages X axis or along the Z axis,as the left two cases shown in relative positions in the 3D space are determined. Fig.11,we can determine their orders along their lining up Taking the scene in Fig.12 for example,there are eight direction by comparing their perpendicular distances of their packages in total (we use the packages'centers to represent centers.For instance,the perpendicular distance of package the corresponding packages).The dash lines are parallel to A should be smaller than others,so package A is ahead of the different axes respectively.The antenna is above all tags. other packages along the X/Z axis.If not,however,as the It performs a 2-dimensional mobile scanning along the X 2 x 2 package stacking in Fig.11,we cannot identify the and Y axes.All of the tags are always at the same side orders of the packages exactly along the X and Z axes at the against the antenna plane.For the scanning along the X axis, same time only through the 1-dimensional mobile scanning. based on the perpendicular points of the packages'centers, To solve this problem,our solution is to perform one more the packages can be split into two sets:[P2,P3,P6,Pi mobile scanning along the orthogonal direction of the previous and [P,P,P,Ps).In each set,the X coordinates of the scanning direction,so as to limit the number of the free packages are the same,as they share the same perpendicular dimensions in the 3D space.Note that,the more times of the points projected to the X axis.Similarly,based on the mobile mobile scanning is,the much more the cost will be,thus we scanning along the Y axis,the packages can also be split into adopt the mobile scanning twice as the least needed times. two sets:[P5,P6,P7,Ps}and [P1,P2,P3,P}.Combining these two split results,we have four sets then:[P,P, B.Determine Package Stacking with a 2-dimensional Mobile (P2,P,(Ps,Ps}and {P6,P}(in Fig.13).In each set Scanning the packages share the same coordinates along the X and Y Through once mobile scanning,after determining the pack- axes.Then,we just need to determine the packages'orders age orientation for each single package,we derive the indica- along the Z axis.By referring to the perpendicular distances tors of the packages'centers for the relative positions along of the packages'centers,we can identify the packages'orders the scanning direction,so we know their relative orders along in each piece.As we determine the relative positions of each that direction.Similarly,by performing the other scanning piece,and the packages'relative positions in each piece,the along the orthogonal direction of the previous one,we can stacking situation of these packages are determined,the 3D get the packages'relative positions along the new direction. reconstruction for multiple packages is done

ο݀௩ ο݀௛ ܶଵ ܶଶ ܶଷ Fig. 10. The distance differences are different for the equal tag spacing along different directions ܣ ܤ ܥ ܦ ܣ ܤ ܥ ܦ ܣ ܤ ܥ ܦ    Fig. 11. Package stacking ‘‹‰ †‹‡…‹‘    ଼ܲ ܲଵ ܲ ܲ଺ ହ ܲସ ܲଶ ܲଷ ܲ଻ Fig. 12. Determining package stacking in the 3D space VII. DETERMINE PACKAGE STACKING FOR MULTIPLE PACKAGES A. Limitations of the 1-dimensional Mobile Scanning for Determining Package Stacking When we determine the package orientation, we derive the indicators for the relative positions among the tags on a single package. As the geometry relationships of these tags are known, we can combine these indicators (perpendicular dis￾tance and perpendicular point) from different tags to estimate the indicators of the package’s center point. Then, similarly, we compare the indicators of the center points of different packages to determine their stacking situation. Note that, when we determine the package orientation, we only need to perform 1-dimensional scanning, but it may not support the package stacking determination due to the 3-DoF in the 3D space. As shown in Fig. 11, suppose the antenna is above all the packages and is in front of the packages along the X axis, it moves along the Y axis. It is easy to determine the packages’ orders along the Y axis by referring to their perpendicular points, but it may be a problem to determine their orders in the XZ plane. If the packages line up , that is, the packages are along the X axis or along the Z axis, as the left two cases shown in Fig. 11, we can determine their orders along their lining up direction by comparing their perpendicular distances of their centers. For instance, the perpendicular distance of package A should be smaller than others, so package A is ahead of other packages along the X/Z axis. If not, however, as the 2 × 2 package stacking in Fig. 11, we cannot identify the orders of the packages exactly along the X and Z axes at the same time only through the 1-dimensional mobile scanning. To solve this problem, our solution is to perform one more mobile scanning along the orthogonal direction of the previous scanning direction, so as to limit the number of the free dimensions in the 3D space. Note that, the more times of the mobile scanning is, the much more the cost will be, thus we adopt the mobile scanning twice as the least needed times. B. Determine Package Stacking with a 2-dimensional Mobile Scanning Through once mobile scanning, after determining the pack￾age orientation for each single package, we derive the indica￾tors of the packages’ centers for the relative positions along the scanning direction, so we know their relative orders along that direction. Similarly, by performing the other scanning along the orthogonal direction of the previous one, we can get the packages’ relative positions along the new direction.   ܲଶ ܲଷ ܲ଺ ܲ଻ ܲଵ ܲସ ܲହ ଼ܲ ଶݔ ଵݔ (a) along the X axis   ܲହ ܲ଺ ܲ଻ ଼ܲ ܲଵ ܲଶ ܲଷ ܲସ ଶݕ ଵݕ (b) along the Y axis   ଵݕ ଻ܲ ଺ܲ ݕଶ ܲଶ ܲଷ ܲଵ ܲସ ܲହ ଼ܲ ଶݔ ଵݔ (c) split plane    (d) split space Fig. 13. Illustration of determining the packages’ order in the 3D space Combining their relative positions along the two directions, the 3D space is divided into many pieces. For each piece, two dimensions are fixed, the packages in it are lining up, so we can use the perpendicular distances of the packages’ centers to determine their orders in the piece. Thus, all the packages’ relative positions in the 3D space are determined. Taking the scene in Fig. 12 for example, there are eight packages in total (we use the packages’ centers to represent the corresponding packages). The dash lines are parallel to the different axes respectively. The antenna is above all tags. It performs a 2-dimensional mobile scanning along the X and Y axes. All of the tags are always at the same side against the antenna plane. For the scanning along the X axis, based on the perpendicular points of the packages’ centers, the packages can be split into two sets: {P2, P3, P6, P7} and {P1, P4, P5, P8}. In each set, the X coordinates of the packages are the same, as they share the same perpendicular points projected to the X axis. Similarly, based on the mobile scanning along the Y axis, the packages can also be split into two sets: {P5, P6, P7, P8} and {P1, P2, P3, P4}. Combining these two split results, we have four sets then: {P1, P4}, {P2, P3}, {P5, P8} and {P6, P7} (in Fig. 13). In each set, the packages share the same coordinates along the X and Y axes. Then, we just need to determine the packages’ orders along the Z axis. By referring to the perpendicular distances of the packages’ centers, we can identify the packages’ orders in each piece. As we determine the relative positions of each piece, and the packages’ relative positions in each piece, the stacking situation of these packages are determined, the 3D reconstruction for multiple packages is done.

5后 Both sdes Both sides 08 nele sid Single side 9 0 02 0.30.40.50.60.70.80.9 0.30.40.50.60.70.80.9 Scanning Range(m) Scanning Range(m) Fig.14.Deployment of RF-3DScan Fig.15. Accuracy for different Fig.16.Angle error for different scanning ranges scanning ranges VIII.PERFORMANCE EVALUATION A.Experiment Settings Orientation accuracy comparison with different distances We build a prototype of RF-3DScan,as shown in Fig.14. between the box and the antenna:We adiust the distance Hardware:Our system consists of one ImpinJ Speedway R420 between the antenna and the boxl as 0.8m,Im and 1.2m, reader,one Laird S9028 RFID antenna and multiple ImpinJ and let the antenna perform both sides scanning of 0.5m 40 times for each distance.The results are shown in Fig.19-20 E41-B tags.The antenna is fixed on a moving car.Soffware: We adopt LLRP protocol to communicate with the reader,and As the distance increases from 0.8m to 1.2m,the angle error decreases by 1.84 and the max error is below 4.9.The angle use a special module to control the moving of the car.Our algorithms are implemented in MATLAB and the language errors for the distances of Im and 1.2m are quite similar.For the bottom face accuracy,it decreases as well,but at least Java.Deployment:We let the antenna be above the boxes and 87.5%when the distance is 1.2m. move at a constant speed of 0.12m/s.For diversity,we use three different sizes of boxes.For each box,there are six tags Stacking accuracy:To evaluate the package stacking,we on it as shown in Fig.8(c).the tag spacing of the two tags on change the number of boxes from two to four,put the boxes the same surface is the same.The tag spacings of boxl,box2 along the antenna scanning direction,in front of the antenna and box3 are 23cm.17cm and 20cm. about Im.The boxes are close to each other.The antenna performs scanning from the end of the boxes to the other B.Micro-Benchmarks end.For multiple boxes,as we put them closely and limit the Metrics:To evaluate the package orientation accuracy,we scanning range,the data missing is serious,it is difficult to have two main metrics:bottom face accuracy,and angle error. determine the box's orientation,so the ordering accuracy is The bottom face accuracy is defined as the number of the not high (in Fig.21),around 73%for the case of three boxes. packages whose bottom faces are identified correctly out of the total package number.The angle error is the error between C.Marco-Benchmarks the estimated angle of the vertical faces against the antenna As STPP cannot handle the cases of limited scanning range, plane and the actual angle.For the package stacking.we use we compare RF-3DScan with STPP in the orientation accura- the metric ordering accuracy.The box is ordered correctly cies of different boxes and different box-antenna distances. only when its detected order is the same with the actual order. Different boxes:We randomly chooses box1,box2 or box3, Orientation accuracy comparison with different scanning and let the antenna be in front of the box about Im,perform ranges:The most advantage of our approach compared to both sides scanning of 0.5m 120 times.As shown in Fig.22.the STPP is that we do not require large range scanning,we adjust bottom face accuracy of STPP is about 81.7%,and RF-3DScan the scanning ranges from 0.3m to 0.9m on a single side or both achieves the accuracy about 92.5%,slightly outperforming sides.Taking the scanning range of 0.3m for example,in terms STPP by x1.13.According to the CDF of the angle error of a single side,it means the antenna starts scanning from the as shown in Fig.23,RF-3DScan performs better than STPP, box and moves 0.3m.while for both sides,the scanning range as the median angle error of STPP is about 3.58 and that of is 0.6m,centered on the box.The results are shown in Fig.15- RF-3DScan is 2.52. 16.From the results,we find RF-3DScan performs well as it Different distances:We choose boxl and let the antenna per- achieves the average bottom face accuracies about 95%for form both sides scanning of 0.5m 120 times.The distances are both sides scanning.and about 70%for one side scanning selected within [0.8m,1.2m]randomly.As shown in Fig.22. range of 0.7m.While the accuracy of the one side scanning is the bottom face accuracy of STPP is about 82.5%,while RF- not that good,STPP cannot deal with such limited scanning 3DScan achieves the accuracy about 93.3%,outperforming ranges. STPP by x1.13.Fig.24 shows the CDF of the angle error, Orientation accuracy comparison with different boxes:We from the figure,RF-3DScan still performs better than STPP, put the box in front of the antenna plane about Im,let the as its median error is about 2.13 and STPP's is 3.62 antenna perform both sides scanning of 0.5m 40 times for each Overall,our experimental results show that RF-3DScan box.As shown in Fig.17-18,the bottom face accuracy is above scales better than STPP for different boxes and box-antenna 87%,and the average angle error is below 4.4.Compared distances,as the average bottom face accuracies of RF-3DScan with boxl and box3,box2 has less accuracy due to its smallest and STPP are about 92.5%and 82.5%,while the average angle tag spacing. errors of them are about 4.08 and 5.05 separately

Fig. 14. Deployment of RF-3DScan VIII. PERFORMANCE EVALUATION A. Experiment Settings We build a prototype of RF-3DScan, as shown in Fig.14. Hardware: Our system consists of one ImpinJ Speedway R420 reader, one Laird S9028 RFID antenna and multiple ImpinJ E41-B tags. The antenna is fixed on a moving car. Software: We adopt LLRP protocol to communicate with the reader, and use a special module to control the moving of the car. Our algorithms are implemented in MATLAB and the language Java. Deployment: We let the antenna be above the boxes and move at a constant speed of 0.12m/s. For diversity, we use three different sizes of boxes. For each box, there are six tags on it as shown in Fig. 8(c), the tag spacing of the two tags on the same surface is the same. The tag spacings of box1, box2 and box3 are 23cm, 17cm and 20cm. B. Micro-Benchmarks Metrics: To evaluate the package orientation accuracy, we have two main metrics: bottom face accuracy, and angle error. The bottom face accuracy is defined as the number of the packages whose bottom faces are identified correctly out of the total package number. The angle error is the error between the estimated angle of the vertical faces against the antenna plane and the actual angle. For the package stacking, we use the metric ordering accuracy. The box is ordered correctly only when its detected order is the same with the actual order. Orientation accuracy comparison with different scanning ranges: The most advantage of our approach compared to STPP is that we do not require large range scanning, we adjust the scanning ranges from 0.3m to 0.9m on a single side or both sides. Taking the scanning range of 0.3m for example, in terms of a single side, it means the antenna starts scanning from the box and moves 0.3m. while for both sides, the scanning range is 0.6m, centered on the box. The results are shown in Fig.15- 16. From the results, we find RF-3DScan performs well as it achieves the average bottom face accuracies about 95% for both sides scanning, and about 70% for one side scanning range of 0.7m. While the accuracy of the one side scanning is not that good, STPP cannot deal with such limited scanning ranges. Orientation accuracy comparison with different boxes: We put the box in front of the antenna plane about 1m, let the antenna perform both sides scanning of 0.5m 40 times for each box. As shown in Fig.17-18, the bottom face accuracy is above 87%, and the average angle error is below 4.4◦. Compared with box1 and box3, box2 has less accuracy due to its smallest tag spacing. 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Scanning Range (m) 0 0.2 0.4 0.6 0.8 1 Accuracy Both sides Single side Fig. 15. Accuracy for different scanning ranges 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Scanning Range (m) 0 5 10 15 Angle Error (deg.) Both sides Single side Fig. 16. Angle error for different scanning ranges Orientation accuracy comparison with different distances between the box and the antenna: We adjust the distance between the antenna and the box1 as 0.8m, 1m and 1.2m, and let the antenna perform both sides scanning of 0.5m 40 times for each distance. The results are shown in Fig.19-20. As the distance increases from 0.8m to 1.2m, the angle error decreases by 1.84◦ and the max error is below 4.9◦. The angle errors for the distances of 1m and 1.2m are quite similar. For the bottom face accuracy, it decreases as well, but at least 87.5% when the distance is 1.2m. Stacking accuracy: To evaluate the package stacking, we change the number of boxes from two to four, put the boxes along the antenna scanning direction, in front of the antenna about 1m. The boxes are close to each other. The antenna performs scanning from the end of the boxes to the other end. For multiple boxes, as we put them closely and limit the scanning range, the data missing is serious, it is difficult to determine the box’s orientation, so the ordering accuracy is not high (in Fig. 21), around 73% for the case of three boxes. C. Marco-Benchmarks As STPP cannot handle the cases of limited scanning range, we compare RF-3DScan with STPP in the orientation accura￾cies of different boxes and different box-antenna distances. Different boxes: We randomly chooses box1, box2 or box3, and let the antenna be in front of the box about 1m, perform both sides scanning of 0.5m 120 times. As shown in Fig.22, the bottom face accuracy of STPP is about 81.7%, and RF-3DScan achieves the accuracy about 92.5%, slightly outperforming STPP by ×1.13. According to the CDF of the angle error as shown in Fig. 23, RF-3DScan performs better than STPP, as the median angle error of STPP is about 3.58◦ and that of RF-3DScan is 2.52◦. Different distances: We choose box1 and let the antenna per￾form both sides scanning of 0.5m 120 times. The distances are selected within [0.8m, 1.2m] randomly. As shown in Fig. 22, the bottom face accuracy of STPP is about 82.5%, while RF- 3DScan achieves the accuracy about 93.3%, outperforming STPP by ×1.13. Fig. 24 shows the CDF of the angle error, from the figure, RF-3DScan still performs better than STPP, as its median error is about 2.13◦ and STPP’s is 3.62◦. Overall, our experimental results show that RF-3DScan scales better than STPP for different boxes and box-antenna distances, as the average bottom face accuracies of RF-3DScan and STPP are about 92.5% and 82.5%, while the average angle errors of them are about 4.08◦ and 5.05◦ separately

0.8 0.8 0 04 02 01 2 0.8 12 0.8 2 Box Box Distance(m) Distance(m Fig.17. Accuracy for different Fig.18.Angle error for different Fig.19. Accuracy for different Fig.20. Angle error for different boxes boxes distances distances 0 0.8 0. 0.6 0.6 0.4 0.2 0.2 -RF-3DScan 0.2 -RF-3DScan 02 F-3DSea -STPP -STPP 0 101520 25 8 12 Number of Boxes Box Distance Angle Error (deg.) Angle Error (deg.) Fig.21. Accuracy for package Fig.22. Accuracy for different Fig.23.CDF of angle error for Fig.24.CDF of angle error for stacking cases different boxes different distances IX.CONCLUSION [6]T.Wei and X.Zhang.Gyro in the Air:Tracking 3D Orientation of In this paper,we present RF-3DScan,an RFID-based system Batteryless Internet-of-Things,in Proc.of ACM MOBICOM,2016. to perform 3D reconstruction on tagged packages.RF-3DScan [7]J.Wang and D.Katabi,Dude.where's my card?:RFID Positioning That Works with Multipath and Non-Line of Sight.in Proc.of ACM can determine the package orientation for a single package SIGCOMM,2013. with the 1-dimensional mobile scanning,and determine the [8]L.Yang.Q.Lin,X.Li.C.Xiao,M.Li and Y.Liu,See Through Walls with COTS RFID Systems,in Proc.of ACM MOBICOM,2015. package stacking for multiple packages with the 2-dimensional [9]W.Ruan.L.Yao.Q.Sheng.N.Falkner and X.Li.TagTrack:Device-free mobile scanning.The key innovation of this work is that we Localization and Tracking Using Passive RFID Tags,in Proc.of ACM propose an angle-profile-based measurement for the relative MOBIOUITOUS.2014. [10]L.Yang.Y.Chen,X.Li,C.Xiao,M.Li,and Y.Liu,Tagoram:Real-time localization.and we show the solution to use the relative Tracking of Mobile RFID Tags to High Precision Using COTS Devices, positions of the tags on the packages to reconstruct the in Proc.of ACM MOBICOM,2014. packages in the 3D space.In the future,we will further [11]J.Wang.F.Adib,R.Knepper,D.Katabi and D.Rus,RF-Compass: Robot Object Manipulation Using RFIDs,in Proc.of ACM MOB/COM. improve our approach,and we wish our work can benefit the 2013. logistic-related applications. [12]J.Wang,D.Vasisht and D.Katabi,RF-IDraw:Virtual Touch Screen in the Air Using RF Signals,in Proc.of ACM S/GCOMM,2014. ACKNOWLEDGMENT (13]L.Shangguan,Z.Yang.A.X.Liu,Z.Zhou and Y.Liu,STPP: This work is supported in part by National Natural Science Spatial-Temporal Phase Profiling-Based Method for Relative RFID Tag Localization,in IEEE/ACM Transactions on Networking (ToN),vol.25 Foundation of China under Grant Nos.61472185.61373129. no.1,Pp.596-609,2017. 61321491,61502224;JiangSu Natural Science Foundation [14]L.Xie,J.Sun,Q.Cai.C.Wang.J.Wu and S.Lu,Tell Me What I under Grant No.BK20151390.This work is partially sup- See:Recognize RFID Tagged Objects in Augmented Reality Systems,in ported by Collaborative Innovation Center of Novel Software Proc.of ACM UBICOMP.2016. [15]J.Nickels,P.Knierim,B.Koenings,F.Schaub,B.Wiedersheim, Technology and Industrialization.Lei Xie is the corresponding S.Musiol,and M.Weber,Find My Stuff:Supporting Physical Objects author. Search with Relative Positioning.in Proc.of ACM MOBIOUITOUS.2013 [16]J.Liu,B.Xiao,S.Chen,F.Zhu and L.Chen,Fast RFID grouping REFERENCES protocols,in Proc.of IEEE INFOCOM,2015. [1]M.Firman,O.M.Aodha.S.Julier and G.J.Brostow,Structured [17]X.Liu,B.Xiao,S.Zhang and K.Bu,Unknown Tag Identification Prediction of Unobserved Voxels From a Single Depth Image,in Proc. in Large RFID Systems:An Efficient and Complete Solution,/EEE of IEEE CVPR,2016. Transactions on Parallel and Distributed System,vol.26,no.6,pp.1775- [2]S.Izadi.D.Kim.O.Hilliges.D.Molyneaux.R.Newcombe.P.Kohli. 1788,2015. J.Shotton,S.Hodges,D.Freeman,A.Davison and A.Fitzgibbon. [18]X.Liu,S.Zhang,B.Xiao and K.Bu,Flexible and Time-Efficient KinectFusion:Real-time 3D Reconstruction and Interaction Using a Tag Scanning with Handheld Readers,IEEE Transactions on Mobile Moving Depth Camera,in Proc.of ACM UIST.2011. Computing.vol.15.no.4.pp.840-852.2016. [3]P.Tanskanen,K.Kolev,L.Meier,F.Camposeco,O.Saurer and M.Polle- [19]J.Han,Q.Chen,X.Wang.D.Ma,J.Zhao,W.Xi,Z.Jiang and Z.Wang feys,Live Metric 3D Reconstruction on Mobile Phones,in Proc.of ACM Twins:Device-free Object Tracking using Passive Tags,in Proc.of IEEE CVPR.2013. INFOCOM.2014. [4]C.Holenstein,R.Zlot and M.Bosse.Watertight Surface Reconstruction [20]L.Yang.J.Han,Y.Qi,C.Wang.T.Gu and Y.Liu,Season:Shelving of Caves from 3D Laser Data,in Proc.of IEEE IROS,2011. interference and joint identification in large-scale rfid systems.in Proc [5]Q.Lin,L.Yang.Y.Sun,T.Liu,X.Li,and Y.Liu,Beyond One-dollar of IEEE INFOCOM.2011. Mouse:A Battery-free Device for 3D Human-computer Interaction via RFID tags,in Proc.of IEEE INFOCOM,2015

123 Box 0 0.2 0.4 0.6 0.8 1 Accuracy Fig. 17. Accuracy for different boxes 123 Box 0 2 4 6 8 Angle Error (deg.) Fig. 18. Angle error for different boxes 0.8 1 1.2 Distance (m) 0 0.2 0.4 0.6 0.8 1 Accuracy Fig. 19. Accuracy for different distances 0.8 1 1.2 Distance (m) 0 2 4 6 8 Angle Error (deg.) Fig. 20. Angle error for different distances Fig. 21. Accuracy for package stacking Box Distance 0 0.2 0.4 0.6 0.8 1 Accuracy RF-3DScan STPP Fig. 22. Accuracy for different cases 0 5 10 15 20 25 Angle Error (deg.) 0 0.2 0.4 0.6 0.8 1 CDF RF-3DScan STPP Fig. 23. CDF of angle error for different boxes 0 4 8 12 16 Angle Error (deg.) 0 0.2 0.4 0.6 0.8 1 CDF RF-3DScan STPP Fig. 24. CDF of angle error for different distances IX. CONCLUSION In this paper, we present RF-3DScan, an RFID-based system to perform 3D reconstruction on tagged packages. RF-3DScan can determine the package orientation for a single package with the 1-dimensional mobile scanning, and determine the package stacking for multiple packages with the 2-dimensional mobile scanning. The key innovation of this work is that we propose an angle-profile-based measurement for the relative localization, and we show the solution to use the relative positions of the tags on the packages to reconstruct the packages in the 3D space. In the future, we will further improve our approach, and we wish our work can benefit the logistic-related applications. ACKNOWLEDGMENT This work is supported in part by National Natural Science Foundation of China under Grant Nos. 61472185, 61373129, 61321491, 61502224; JiangSu Natural Science Foundation under Grant No. BK20151390. This work is partially sup￾ported by Collaborative Innovation Center of Novel Software Technology and Industrialization. Lei Xie is the corresponding author. REFERENCES [1] M. Firman, O. M. Aodha, S. Julier and G. J. Brostow, Structured Prediction of Unobserved Voxels From a Single Depth Image, in Proc. of IEEE CVPR, 2016. [2] S. Izadi, D. Kim, O. Hilliges, D. Molyneaux, R. Newcombe, P. Kohli, J. Shotton, S. Hodges, D. Freeman, A. Davison and A. Fitzgibbon, KinectFusion: Real-time 3D Reconstruction and Interaction Using a Moving Depth Camera, in Proc. of ACM UIST, 2011. [3] P. Tanskanen, K. Kolev, L. Meier, F. Camposeco, O. Saurer and M. Polle￾feys, Live Metric 3D Reconstruction on Mobile Phones, in Proc. of ACM CVPR, 2013. [4] C. Holenstein, R. Zlot and M. Bosse, Watertight Surface Reconstruction of Caves from 3D Laser Data, in Proc. of IEEE IROS, 2011. [5] Q. Lin, L. Yang, Y. Sun, T. Liu, X. Li, and Y. Liu, Beyond One-dollar Mouse: A Battery-free Device for 3D Human-computer Interaction via RFID tags, in Proc. of IEEE INFOCOM, 2015. [6] T. Wei and X. Zhang, Gyro in the Air: Tracking 3D Orientation of Batteryless Internet-of-Things, in Proc. of ACM MOBICOM, 2016. [7] J. Wang and D.Katabi, Dude, where’s my card?: RFID Positioning That Works with Multipath and Non-Line of Sight, in Proc. of ACM SIGCOMM, 2013. [8] L. Yang, Q. Lin, X. Li, C. Xiao, M. Li and Y. Liu, See Through Walls with COTS RFID Systems, in Proc. of ACM MOBICOM, 2015. [9] W. Ruan, L. Yao, Q. Sheng, N. Falkner and X. Li, TagTrack: Device-free Localization and Tracking Using Passive RFID Tags, in Proc. of ACM MOBIQUITOUS, 2014. [10] L. Yang, Y. Chen, X. Li, C. Xiao, M. Li, and Y. Liu, Tagoram: Real-time Tracking of Mobile RFID Tags to High Precision Using COTS Devices, in Proc. of ACM MOBICOM, 2014. [11] J. Wang, F. Adib, R. Knepper, D. Katabi and D. Rus, RF-Compass: Robot Object Manipulation Using RFIDs, in Proc. of ACM MOBICOM, 2013. [12] J. Wang, D. Vasisht and D. Katabi, RF-IDraw: Virtual Touch Screen in the Air Using RF Signals, in Proc. of ACM SIGCOMM, 2014. [13] L. Shangguan, Z. Yang, A. X. Liu, Z. Zhou and Y. Liu, STPP: Spatial-Temporal Phase Profiling-Based Method for Relative RFID Tag Localization, in IEEE/ACM Transactions on Networking (ToN), vol. 25, no. 1, pp. 596-609, 2017. [14] L. Xie, J. Sun, Q. Cai, C. Wang, J. Wu and S. Lu, Tell Me What I See: Recognize RFID Tagged Objects in Augmented Reality Systems, in Proc. of ACM UBICOMP, 2016. [15] J. Nickels, P. Knierim, B. Koenings, F. Schaub, B. Wiedersheim, S. Musiol, and M. Weber, Find My Stuff: Supporting Physical Objects Search with Relative Positioning, in Proc. of ACM MOBIQUITOUS, 2013. [16] J. Liu, B. Xiao, S. Chen, F. Zhu and L. Chen, Fast RFID grouping protocols, in Proc. of IEEE INFOCOM, 2015. [17] X. Liu, B. Xiao, S. Zhang and K. Bu, Unknown Tag Identification in Large RFID Systems: An Efficient and Complete Solution, IEEE Transactions on Parallel and Distributed System, vol. 26, no. 6, pp. 1775- 1788, 2015. [18] X. Liu, S. Zhang, B. Xiao and K. Bu, Flexible and Time-Efficient Tag Scanning with Handheld Readers, IEEE Transactions on Mobile Computing, vol. 15, no. 4, pp. 840-852, 2016. [19] J. Han, Q. Chen, X. Wang, D. Ma, J. Zhao, W. Xi, Z. Jiang and Z. Wang, Twins: Device-free Object Tracking using Passive Tags, in Proc. of IEEE INFOCOM, 2014. [20] L. Yang, J. Han, Y. Qi, C. Wang, T. Gu and Y. Liu, Season: Shelving interference and joint identification in large-scale rfid systems, in Proc. of IEEE INFOCOM, 2011