Exoskeletons for Human Performance Augmentation 33.7 The Control Scheme of an Exoskeleton 783 The exoskeleton velocity,as shown by (33.9),is from,for example,the human hands.and measur- affected by forces and torques from its pilot.The sensi- ing forces from the human lower limbs.Using our tivity transfer function,S,represents how the equivalent hands,we are able to impose controlled forces and human torque affects the exoskeleton angular velocity; torques on upper extremity exoskeletons and haptic S maps the equivalent pilot torque d onto the exoskel- systems with very few uncertainties.However,our eton velocity v.If the actuator already has some sort of lower limbs have other primary and nonvoluntary primary stabilizing controller,the magnitude of S will functions like load support that take priority over be small and the exoskeleton will only have a small re- locomotion. sponse to the imposed forces and torques from the pilot 3.One option which was experimented with was the or any other source.For example,a high-gain velocity installation of sensing devices for forces on the bot- controller in the actuator results in small S,and con- tom of the pilot's boots,where they are connected to sequently a small exoskeleton response to forces and the exoskeleton.Since the force on the bottom of the torques.Also,non-back-drivable actuators (e.g.,large pilot's boot travels from heel to toe during normal transmission ratios or servo-valves with overlapping walking,several sensors are required to measure the spools)result in a small S,which leads to a correspond- pilot force.Ideally,one would have a matrix of force ingly small response to pilot forces and torques. sensors between the pilot and exoskeleton feet to Note that d (the resulting torque from pilot on the measure the pilot forces at all locations and at all di- exoskeleton)is not an exogenous input;it is a function rections,though in practice,only a few sensors could of the pilot dynamics and variables such as position and be accommodated:at the toe,ball,midfoot,and the velocity of the pilot and the exoskeleton legs.These heel.Still,this option leads to thick and bulky soles. dynamics change from person to person,and within 4.The bottoms of the human boots experience cyclic a person as a function of time and posture.It is as- forces and torques during normal walking that lead sumed that d is only from the pilot and does not include to fatigue and eventual sensor failure if the sensor is any other external forces and torques. not designed and isolated properly. The objective is to increase exoskeleton sensitivity to pilot forces and torques through feedback but without For the above reasons and our experience in the measuring d.Measuring d to create such systems devel- design of various lower-extremity exoskeletons,it be- ops several hard,but ultimately solvable,problems in came evident that the existing state of technology in the control of a lower-extremity exoskeleton.Some of force sensing could not provide robust and repeatable those problems are briefly described below. measurement of the human lower limb force on the exoskeleton.Our goal then shifted to developing an exo- 1.Depending on the architecture and the design of the skeleton with a large sensitivity to forces and torques exoskeleton,one needs to install several force and from the operator using measurements only from the torque sensors to measure all forces from the pilot on exoskeleton (i.e.,no sensors on the pilot or the exo- the exoskeleton because the pilot is in contact with skeleton interface with the pilot).Creating a feedback the exoskeleton at several locations.These locations loop from the exoskeleton variables only,as shown are not known in advance.For example,we have in Fig.33.13 the new closed-loop sensitivity transfer found that some pilots are interested in having braces function is presented in (33.10). connecting an exoskeleton at the shanks while some are interested in having them on the thighs.Inclusion S SNEW=一= (33.10) of sensors on a leg to measure all kinds of human d 1+GC forces and torques may result in a system suitable for a laboratory setting but not robust enough to be deployed in the field. 2.If the exoskeleton design is such that the forces and torques applied by the pilot on the exoskeleton are limited to a specified location (e.g.,the pilot Part foot),the sensor that measures the pilot forces and torques will also inadvertently measure other forces Fig.33.13 The feedback control loop added to block dia- 出 and torques that are not intended for locomotion. gram of Fig.33.12;C is the controller operating only on This is a major difference between measuring forces the exoskeleton variablesExoskeletons for Human Performance Augmentation 33.7 The Control Scheme of an Exoskeleton 783 The exoskeleton velocity, as shown by (33.9), is affected by forces and torques from its pilot. The sensi￾tivity transfer function, S, represents how the equivalent human torque affects the exoskeleton angular velocity; S maps the equivalent pilot torque d onto the exoskel￾eton velocity v. If the actuator already has some sort of primary stabilizing controller, the magnitude of S will be small and the exoskeleton will only have a small re￾sponse to the imposed forces and torques from the pilot or any other source. For example, a high-gain velocity controller in the actuator results in small S, and con￾sequently a small exoskeleton response to forces and torques. Also, non-back-drivable actuators (e.g., large transmission ratios or servo-valves with overlapping spools) result in a small S, which leads to a correspond￾ingly small response to pilot forces and torques. Note that d (the resulting torque from pilot on the exoskeleton) is not an exogenous input; it is a function of the pilot dynamics and variables such as position and velocity of the pilot and the exoskeleton legs. These dynamics change from person to person, and within a person as a function of time and posture. It is as￾sumed that d is only from the pilot and does not include any other external forces and torques. The objective is to increase exoskeleton sensitivity to pilot forces and torques through feedback but without measuring d. Measuring d to create such systems devel￾ops several hard, but ultimately solvable, problems in the control of a lower-extremity exoskeleton. Some of those problems are briefly described below. 1. Depending on the architecture and the design of the exoskeleton, one needs to install several force and torque sensors to measure all forces from the pilot on the exoskeleton because the pilot is in contact with the exoskeleton at several locations. These locations are not known in advance. For example, we have found that some pilots are interested in having braces connecting an exoskeleton at the shanks while some are interested in having them on the thighs. Inclusion of sensors on a leg to measure all kinds of human forces and torques may result in a system suitable for a laboratory setting but not robust enough to be deployed in the field. 2. If the exoskeleton design is such that the forces and torques applied by the pilot on the exoskeleton are limited to a specified location (e.g., the pilot foot), the sensor that measures the pilot forces and torques will also inadvertently measure other forces and torques that are not intended for locomotion. This is a major difference between measuring forces from, for example, the human hands, and measur￾ing forces from the human lower limbs. Using our hands, we are able to impose controlled forces and torques on upper extremity exoskeletons and haptic systems with very few uncertainties. However, our lower limbs have other primary and nonvoluntary functions like load support that take priority over locomotion. 3. One option which was experimented with was the installation of sensing devices for forces on the bot￾tom of the pilot’s boots, where they are connected to the exoskeleton. Since the force on the bottom of the pilot’s boot travels from heel to toe during normal walking, several sensors are required to measure the pilot force. Ideally, one would have a matrix of force sensors between the pilot and exoskeleton feet to measure the pilot forces at all locations and at all di￾rections, though in practice, only a few sensors could be accommodated: at the toe, ball, midfoot, and the heel. Still, this option leads to thick and bulky soles. 4. The bottoms of the human boots experience cyclic forces and torques during normal walking that lead to fatigue and eventual sensor failure if the sensor is not designed and isolated properly. For the above reasons and our experience in the design of various lower-extremity exoskeletons, it be￾came evident that the existing state of technology in force sensing could not provide robust and repeatable measurement of the human lower limb force on the exoskeleton. Our goal then shifted to developing an exo￾skeleton with a large sensitivity to forces and torques from the operator using measurements only from the exoskeleton (i. e., no sensors on the pilot or the exo￾skeleton interface with the pilot). Creating a feedback loop from the exoskeleton variables only, as shown in Fig. 33.13 the new closed-loop sensitivity transfer function is presented in (33.10). SNEW = v d = S 1+ GC . (33.10) S υ d + – G C Fig. 33.13 The feedback control loop added to block dia￾gram of Fig. 33.12; C is the controller operating only on the exoskeleton variables Part D 33.7