Swift augmented human–robot dynamics modeling for rehabilitation planning analyses

Abstract

With the widespread implementation of robotics exoskeletons in rehabilitation, modeling and dynamics analysis of such highly nonlinear coupled systems has become significantly important. In this paper, a swift numerical human–robot dynamics modeling has been developed to achieve accurate and realistic interpretation. This takes into consideration the separation and impact between multiple bodies for rehabilitation planning. To this end, first, a novel parallel algorithm combined with sequential interaction conditions is proposed based on the numerical recursive Newton–Euler method. The approach begins by deriving separated numerical models for the complicated system: i.e. both the human and the robot. These models are then augmented, with a primary focus on reducing the error of the interaction conditions, including forces and positions. The accuracy of the proposed model, with a computational complexity of O(n), is assessed by comparing to a previously validated nonrecursive analytical model with a higher computational complexity of O(n^4). Additionally, the quality of the connection between the human and the robot is assessed to establish a suitable control objective and an effective interaction strategy for rehabilitation planning. The study employs a lower-limb walking assistive robot developed in the ARAS lab (RoboWalk) to validate the proposed method. The algorithm is empirically implemented on the RoboWalk test stand, ensuring the integrity of the proposed dynamics modeling. The human–robot interaction forces are estimated with an accuracy of 2 N, in the presence of friction and measurement noise. Finally, the effectiveness of the model-based controller is assessed by using the proposed method, providing valuable tools for the enhancement of overall performance of such a complex dynamics system.