Vision Assisted Control of Lower Extremity Exoskeleton for Obstacle Avoidance With Dynamic Constraint Based Piecewise Nonlinear MPC

This article proposed a humanoid obstacle passability strategy (OPS) to enhance human-exoskeleton integrated system to cross over multi-obstacle in sagittal plane. A hybrid bounding box integrated with closeness regression in L-shape section and synchronous convergence in convex hull search (HBB-LC) is designed to recognize geometric features of staggered obstacles with partial surface occluded. A human-in-the-loop gait pattern planning mechanism based on nonlinear model predictive control (NMPC) is designed for decoupling mapping. To overcome local oscillation and achieve humanoid behavior transition during NMPC optimization, an intermediate switching criterion based on instantaneous capture point (ICP) transfer principle and guiding interpolation inflection is established to trigger the external intervention of piecewise NMPC models. Exploitation of OPS achieves the contributions below: 1) Sectional attitude and dimension features of spatial multi-obstacle are accurately extracted, with size deviation within 3.4-16.1 mm, recognition accuracy within 89.2%-96.8%. 2) Traversal avoidance region of swing toe/heel orbit (SHO/STO) is enhanced by 28.7%-86.8% (with uplift/sidesway range (UR/SR) by 0.071-0.154 m/0.057-0.092 m). 3) Security margin between STO/SHO and obstacle vertex is improved by 0.05-0.19 m (53.3%-161.1%). 4) Elapsed time of crossing obstacles is reduced by 0.62-1.03 s (with gait phase by 36.2%-43.5%). Experimental verification proves significant augmentation for crossing obstacles by OPS.