Rehabilitation and Health Care Robotics 53.2 Physical Therapy and Training Robots 1233 Of these three robot systems,the GT-I is the one with the Lokomat in 20 chronic incomplete SCI pa- that departs most from therapist-assisted BWSTT,since tients.The improvements in overground walking speed it interacts with the patient's lower limbs through two and endurance were statistically significant:approxi- footplates rather than acting on the shank as human ther- mately 50%gain on average in the 16 patients who were apists do.It also appears to depart more from natural ambulatory before training.There were no significant walking because the footplate principle substantially al- changes in the requirement of walking aids,orthoses ters the sensory cues of the foot impact with the ground or external physical assistance.The improvements ap- or treadmill band.The GT-I footplates are driven by pear to be comparable to those achieved by similar a singly actuated mechanism that moves the foot along SCI patients who received therapist-assisted BWSTT. a fixed gait-like trajectory with a doubled crank and Hornby and colleagues [53.76]studied the effects of rocker system [53.81].The stride length can be ad- robot-assisted BWSTT with the Lokomat on individu- justed between sessions by changing gears.The body als with subacute SCI.Thirty patients were randomly weight is unloaded as needed by an overhead harness. assigned to one of three training groups:robot-assisted The torso is moved sagittally in a phase-dependent man- BWSTT,therapist-assisted BWSTT,and therapist- ner by ropes attached to the harness and connected by assisted overground ambulation.Improvements in motor another crank to the foot crank.The GT-I is currently and functional abilities were similar in the three train- installed in dozens of clinics,mainly in Europe.One ran-ing groups,with the robot-assisted BWSTT requiring domized,controlled study has been reported that tested significantly less labor than the other two therapy the GT-I with 30 subacute stroke patients [53.84].The methods. robot group improved their overground walking abil- The AutoAmbulator(http://www.autoambulator.com) ity more than the control group,although differences consists of two robotic arms that assist patients to step on art were not significant at six-month follow-up.A total of a treadmill with their body weight supported as needed. 80%of the patients said they preferred training with The interface to the patient's legs is through straps at the robot rather than the therapists because training the thigh and ankle.The AutoAmbulator is currently with the robot was less demanding and more comfort- being used in 57 HealthSouth rehabilitation centers,all able.The other 20%of patients stated that swinging of them in the United States.A single-blind,randomized of the paretic limb seemed less natural and thus less clinical trial to assess its effectiveness in stroke patients effective when training with the robot.Robot-assisted is currently underway. training required an average of one therapist per patient, while therapist-assisted training required two thera- Further Research and Development pists per patient on average.A follow-up,randomized on Robotic Therapy for Walking controlled study comparing conventional training plus Several groups worldwide are working toward improv- robotic training with the GT-I with a time-matched ing gait-training robotic technologies.A great deal of amount of conventional training alone with subacute effort has been going into incorporating and investi- stroke patients found that the group that received some gating the ability to assist as needed [53.31,87-91]. robotic training recovered walking ability to a great that is,the ability of the robot to let the patients extent [53.85]. contribute to the locomotor efforts as much as they The Lokomat is a robotic exoskeleton worn by the are able.This is likely essential for maximizing lo- patients during treadmill walking [53.82].Four motor-comotor plasticity [53.92].Some effort has also been ized joints (two per leg)move the hip and knee.The directed towards adding more active DOF,particularly actuators consist of ball screws connected to direct-for torso manipulation [53.90,93].These robotic tools current (DC)motors.The legs are driven in a gait-like are needed not only for their potential clinical use pattern along a fixed position-controlled trajectory.The in therapy,but for studying what aspects of the as- device attaches to the thighs and shanks through padded sistance are important for effective gait training and straps.A passive parallelogram mechanism allows ver-how best to control and implement them with robotic tical translation of the patient's torso,restricting lateral devices. translation.The patient's body weight is unloaded as The team responsible for the GT-1 has developed needed through an overhead harness.The Lokomat is the Haptic Walker [53.87],which maintains the perma- currently being used in dozens of research labs and clin- nent foot/machine contact but allows the footplates to ics worldwide.In 2005,Wirz and coworkers [53.86] move along three-DOF trajectories.In addition it incor- reported preliminary results of robot-assisted BWSTT porates force feedback and compliance control,as wellRehabilitation and Health Care Robotics 53.2 Physical Therapy and Training Robots 1233 Of these three robot systems, the GT-I is the one that departs most from therapist-assisted BWSTT, since it interacts with the patient’s lower limbs through two footplates rather than acting on the shank as human ther￾apists do. It also appears to depart more from natural walking because the footplate principle substantially al￾ters the sensory cues of the foot impact with the ground or treadmill band. The GT-I footplates are driven by a singly actuated mechanism that moves the foot along a fixed gait-like trajectory with a doubled crank and rocker system [53.81]. The stride length can be ad￾justed between sessions by changing gears. The body weight is unloaded as needed by an overhead harness. The torso is moved sagittally in a phase-dependent man￾ner by ropes attached to the harness and connected by another crank to the foot crank. The GT-I is currently installed in dozens of clinics, mainly in Europe. One ran￾domized, controlled study has been reported that tested the GT-I with 30 subacute stroke patients [53.84]. The robot group improved their overground walking abil￾ity more than the control group, although differences were not significant at six-month follow-up. A total of 80% of the patients said they preferred training with the robot rather than the therapists because training with the robot was less demanding and more comfort￾able. The other 20% of patients stated that swinging of the paretic limb seemed less natural and thus less effective when training with the robot. Robot-assisted training required an average of one therapist per patient, while therapist-assisted training required two thera￾pists per patient on average. A follow-up, randomized controlled study comparing conventional training plus robotic training with the GT-I with a time-matched amount of conventional training alone with subacute stroke patients found that the group that received some robotic training recovered walking ability to a great extent [53.85]. The Lokomat is a robotic exoskeleton worn by the patients during treadmill walking [53.82]. Four motor￾ized joints (two per leg) move the hip and knee. The actuators consist of ball screws connected to direct￾current (DC) motors. The legs are driven in a gait-like pattern along a fixed position-controlled trajectory. The device attaches to the thighs and shanks through padded straps. A passive parallelogram mechanism allows ver￾tical translation of the patient’s torso, restricting lateral translation. The patient’s body weight is unloaded as needed through an overhead harness. The Lokomat is currently being used in dozens of research labs and clin￾ics worldwide. In 2005, Wirz and coworkers [53.86] reported preliminary results of robot-assisted BWSTT with the Lokomat in 20 chronic incomplete SCI pa￾tients. The improvements in overground walking speed and endurance were statistically significant: approxi￾mately 50% gain on average in the 16 patients who were ambulatory before training. There were no significant changes in the requirement of walking aids, orthoses or external physical assistance. The improvements ap￾pear to be comparable to those achieved by similar SCI patients who received therapist-assisted BWSTT. Hornby and colleagues [53.76] studied the effects of robot-assisted BWSTT with the Lokomat on individu￾als with subacute SCI. Thirty patients were randomly assigned to one of three training groups: robot-assisted BWSTT, therapist-assisted BWSTT, and therapist￾assisted overground ambulation. Improvements in motor and functional abilities were similar in the three train￾ing groups, with the robot-assisted BWSTT requiring significantly less labor than the other two therapy methods. The AutoAmbulator (http://www.autoambulator.com) consists of two robotic arms that assist patients to step on a treadmill with their body weight supported as needed. The interface to the patient’s legs is through straps at the thigh and ankle. The AutoAmbulator is currently being used in 57 HealthSouth rehabilitation centers, all of them in the United States. A single-blind, randomized clinical trial to assess its effectiveness in stroke patients is currently underway. Further Research and Development on Robotic Therapy for Walking Several groups worldwide are working toward improv￾ing gait-training robotic technologies. A great deal of effort has been going into incorporating and investi￾gating the ability to assist as needed [53.31, 87–91], that is, the ability of the robot to let the patients contribute to the locomotor efforts as much as they are able. This is likely essential for maximizing lo￾comotor plasticity [53.92]. Some effort has also been directed towards adding more active DOF, particularly for torso manipulation [53.90, 93]. These robotic tools are needed not only for their potential clinical use in therapy, but for studying what aspects of the as￾sistance are important for effective gait training and how best to control and implement them with robotic devices. The team responsible for the GT-1 has developed the HapticWalker [53.87], which maintains the perma￾nent foot/machine contact but allows the footplates to move along three-DOF trajectories. In addition it incor￾porates force feedback and compliance control, as well Part F 53.2