UC Irvine

Robotic devices are being developed to automate repetitive aspects of walking retraining after neurological injuries, in part because they might improve the consistency and quality of training. However, it is unclear how inconsistent manual training actually is or whether stepping quality depends strongly on the trainers' manual skill. The objective of this study was to quantify trainer variability of manual skill during step training using body-weight support on a treadmill and assess factors of trainer skill. We attached a sensorized orthosis to one leg of each patient with spinal cord injury and measured the shank kinematics and forces exerted by different trainers during six training sessions. An expert trainer rated the trainers' skill level based on videotape recordings. Between-trainer force variability was substantial, about two times greater than within-trainer variability. Trainer skill rating correlated strongly with two gait features: better knee extension during stance and fewer episodes of toe dragging. Better knee extension correlated directly with larger knee horizontal assistance force, but better toe clearance did not correlate with larger ankle push-up force; rather, it correlated with better knee and hip extension. These results are useful to inform robotic gait-training design.