A new hybrid neurodynamics-based model-less solution for redundant robot fault-tolerant motion planning and control

Abstract

For tasks utilizing redundant manipulators, the motion of multiple joints is involved in performing tracking control. In some cases, the failure of one or more joints may lead to task failure or even cause damage, highlighting the necessity of fault tolerance as a crucial capability for robotic control systems. To achieve the fault-tolerant control capability of the redundant manipulator, a quadratic programming problem is formulated to minimize the joint velocity based on the task-priority strategy. Based on this formulation, a constraint transformation method is employed to handle the joint velocity constraints, and finally, this quadratic programming problem is solved using zeroing neurodynamics with finite-time convergence. Unlike most previous fault-tolerant control algorithms, the proposed method estimates the Jacobian matrix in a data-driven manner based on gradient neurodynamics, without requiring the kinematic model of the redundant manipulator. The effectiveness of the proposed method is evaluated through simulations and experiments using manipulators with different degrees of freedom.