• knee exoskeleton
• Lower limb exoskeleton
• nonlinear low-level controller

The aim of this thesis is the investigation on low-level control strategies for enhancing walking assistance with a nonlinear active knee orthosis (AKO), characterized by intrinsic model-free disturbances. Preliminary tests with a linear torque controller showed that assistance by torque control with the AKO might be suitable for swing phase, but not for stance phase. A new proposed assistive strategy for stance phase was position control for knee stabilization. A linear position controller C(s) was tuned on an AKO model, obtained through system identification, showing poor performances in stabilizing the AKO against unpredictable disturbances. Two nonlinear position control approaches were, thus, proposed. Both multiplied C(s) by a variable gain factor G, function of position error e(t) and its derivative de(t)/dt. Method A assigned different fixed values to G, depending on predetermined thresholds of both e(t) and de(t)/dt, while Method B computed G with an exponential function of them. Method B showed higher robustness than Method A on the AKO, suggesting its possible application for stance assistance. A final hybrid position-torque control assistive strategy was proposed. It delivered assistance during swing with the tested linear torque controller, and with a Method B-based novel nonlinear position controller during stance. Pilot tests on healthy subjects showed, finally, improved performances of this strategy with respect to a strategy based only on torque delivery.