strength of backscatter responses,reduce data rates to enhance robustness against noise,and notify coexisting wireless devices to mitigate interferences.Although our approach cannot mag- ically fight against strong channel variation (due to intentional inference,fast tag mobility,dramatic environment dynamics, etc.),P-MTI is experimentally proven to be robust in rea- Fig.4.Testbed:2 circular antennas are mounted to USRP N210.The USRP sonably good channel conditions (Section V)and stationary N210 is connected via GigE to a laptop which acts as an RFID reader. environment settings (Section IV). E.Multiple readers Multiple readers are normally deployed to ensure a full coverage of a large monitoring area [28]due to power con- straints,interrogation environment,etc.For instance,current Fig.5.A WISP tag. RFID readers can only power up the passive RFID tags within approximately 10m.We denote the number of RFID only perform lightweight routine computation tasks required readers as L which is generally constrained by the deployment by the EPCglobal Gen-2 standard [3],and the computational budget.With more RFID readers,the number of tags in each overhead is negligible. monitoring area can be reduced,which mitigates the tag-tag 3)Channel dynamics:In practice,wireless channels contention and collision in the area.Besides,the RFID readers change over time.P-MTI naturally embraces the channel with non-overlapping coverage can interrogate tags in parallel dynamics.First,the compressive sensing based signal recon- without the reader-reader interference. struction is robust to channel noise.P-MTI takes measurement Recent works propose to efficiently schedule multiple RFID noise (due to channel dynamics,interference,quantization, readers to improve the performance [28].P-MTI is able to etc.)into consideration and enhance its robustness based on adopt similar coordination strategies to achieve the parallel the theory of stable recovery (Section III-D).Second,the interrogation of multiple readers.We note that a tag can be detection function f(zi)obviates the need of accurate chan- in the overlapped coverage of multiple RFID readers.As the nel measurements.In stationary environment settings (e.g., readers with overlapped coverage can easily be scheduled medicine store,military basis,etc),the channel variation is into different time slots by the server,the tags will only talk small.If channel condition changes dramatically during a to one reader at a time.Therefore,each RFID reader with short period,P-MTI may draw a false detection result.One non-overlapping coverage area can interrogate the tags in its possible solution is to do extra measurements to ensure the coverage.Each RFID reader reports all the missing tags as detection results.For instance,the RFID reader may query the well as the present tags in its coverage to the backend server. potential missing tags to respond immediately.If no response The server claims a missing tag event only when a tag is is sent back from the tag,then the reader can confidently absent in all the monitoring area of the RFID readers.Such conclude its absence.Tag mobility also introduces channel a simple data processing task can be easily handled by the dynamics.While P-MTI primarily focus on monitoring static backend server with powerful computation capability. goods (e.g.,in inventory management),our scheme inherently tolerates low mobility scenarios where the channel dynamics F.Discussion are within the stability range as discussed in Section III. 1)Communication cost:The RFID reader needs to initiate Following the existing approaches [16,23],we focus on the the missing tag identification process by sending a command wireless channels without the intentional interference from as well as communication parameters (e.g.,data rate,encoding adversaries.P-MTI may benefit from a wise channel selection scheme.etc).Such an initialization is typically in orders scheme (e.g.,BLINK [30])to ensure the channel quality as of ms.As the RFID readers are normally connected with well. backend server via high speed links,the communication cost between readers and backend server is also small compared IV.IMPLEMENTATION with that of the backscatter communication.The backscatter Although the EPCglobal Gen-2 standard specifies many communication from tags to reader involves the transmission operations of tags and readers,the practical implementation time of O(K log(N/K))physical layer symbols. is left to the manufacturers'choices.Production RFID readers 2)Computational complexity:The computation time of (e.g.,Alien ALR 9900+RFID reader [1])only provide limited compressive sensing based decoding performed at the backend interfaces and do not expose physical layer information to server turns out to be the major contributor to the overall users.To explore the lower layer information,we build a computation overhead.In our implementation,we use the off- prototype missing tag identification system based on the USRP the-shelf CVX solver [2]to decode the aggregated responses software defined radio and the programmable WISP tags. and identify the missing tags,which involves computation time Figure 4 shows the testbed.In particular,we implement a of O(N3).Our implementation using commodity PCs can prototype software defined RFID reader using USRP N210 easily cope with the computation tasks in sub-second time with based on the GNURadio toolkit and Gen2 RFID project [5]. thousands of tags.In P-MTI,both RFID reader and RFID tag One USRP RFX900 daughterboard operating in the 900MHzstrength of backscatter responses, reduce data rates to enhance robustness against noise, and notify coexisting wireless devices to mitigate interferences. Although our approach cannot mag￾ically fight against strong channel variation (due to intentional inference, fast tag mobility, dramatic environment dynamics, etc.), P-MTI is experimentally proven to be robust in rea￾sonably good channel conditions (Section V) and stationary environment settings (Section IV). E. Multiple readers Multiple readers are normally deployed to ensure a full coverage of a large monitoring area [28] due to power con￾straints, interrogation environment, etc. For instance, current RFID readers can only power up the passive RFID tags within approximately 10m. We denote the number of RFID readers as L which is generally constrained by the deployment budget. With more RFID readers, the number of tags in each monitoring area can be reduced, which mitigates the tag-tag contention and collision in the area. Besides, the RFID readers with non-overlapping coverage can interrogate tags in parallel without the reader-reader interference. Recent works propose to efficiently schedule multiple RFID readers to improve the performance [28]. P-MTI is able to adopt similar coordination strategies to achieve the parallel interrogation of multiple readers. We note that a tag can be in the overlapped coverage of multiple RFID readers. As the readers with overlapped coverage can easily be scheduled into different time slots by the server, the tags will only talk to one reader at a time. Therefore, each RFID reader with non-overlapping coverage area can interrogate the tags in its coverage. Each RFID reader reports all the missing tags as well as the present tags in its coverage to the backend server. The server claims a missing tag event only when a tag is absent in all the monitoring area of the RFID readers. Such a simple data processing task can be easily handled by the backend server with powerful computation capability. F. Discussion 1) Communication cost: The RFID reader needs to initiate the missing tag identification process by sending a command as well as communication parameters (e.g., data rate, encoding scheme, etc). Such an initialization is typically in orders of ms. As the RFID readers are normally connected with backend server via high speed links, the communication cost between readers and backend server is also small compared with that of the backscatter communication. The backscatter communication from tags to reader involves the transmission time of O(K log(N/K)) physical layer symbols. 2) Computational complexity: The computation time of compressive sensing based decoding performed at the backend server turns out to be the major contributor to the overall computation overhead. In our implementation, we use the off￾the-shelf CVX solver [2] to decode the aggregated responses and identify the missing tags, which involves computation time of O(N3 ). Our implementation using commodity PCs can easily cope with the computation tasks in sub-second time with thousands of tags. In P-MTI, both RFID reader and RFID tag Fig. 4. Testbed: 2 circular antennas are mounted to USRP N210. The USRP N210 is connected via GigE to a laptop which acts as an RFID reader. Fig. 5. A WISP tag. only perform lightweight routine computation tasks required by the EPCglobal Gen-2 standard [3], and the computational overhead is negligible. 3) Channel dynamics: In practice, wireless channels change over time. P-MTI naturally embraces the channel dynamics. First, the compressive sensing based signal recon￾struction is robust to channel noise. P-MTI takes measurement noise (due to channel dynamics, interference, quantization, etc.) into consideration and enhance its robustness based on the theory of stable recovery (Section III-D). Second, the detection function f(z∆i) obviates the need of accurate chan￾nel measurements. In stationary environment settings (e.g., medicine store, military basis, etc), the channel variation is small. If channel condition changes dramatically during a short period, P-MTI may draw a false detection result. One possible solution is to do extra measurements to ensure the detection results. For instance, the RFID reader may query the potential missing tags to respond immediately. If no response is sent back from the tag, then the reader can confidently conclude its absence. Tag mobility also introduces channel dynamics. While P-MTI primarily focus on monitoring static goods (e.g., in inventory management), our scheme inherently tolerates low mobility scenarios where the channel dynamics are within the stability range as discussed in Section III. Following the existing approaches [16, 23], we focus on the wireless channels without the intentional interference from adversaries. P-MTI may benefit from a wise channel selection scheme (e.g., BLINK [30]) to ensure the channel quality as well. IV. IMPLEMENTATION Although the EPCglobal Gen-2 standard specifies many operations of tags and readers, the practical implementation is left to the manufacturers’ choices. Production RFID readers (e.g., Alien ALR 9900+ RFID reader [1]) only provide limited interfaces and do not expose physical layer information to users. To explore the lower layer information, we build a prototype missing tag identification system based on the USRP software defined radio and the programmable WISP tags. Figure 4 shows the testbed. In particular, we implement a prototype software defined RFID reader using USRP N210 based on the GNURadio toolkit and Gen2 RFID project [5]. One USRP RFX900 daughterboard operating in the 900MHz