Robust Event-Triggered Optimal Control for Lower Limb Exoskeleton Robots With Friction and Unknown Perturbations in an Interactive Environment

In this article, we propose a new adaptive critic neural networks (Critic-NNs) learning algorithm for robust optimal tracking control of nonzero-equilibrium lower limb exoskeleton robots (LLER) with friction and unknown perturbations in an interactive environment. By introducing a nominal system, the robust tracking control of the original system is transformed into an optimal tracking control problem of the nominal system. The traditional adaptive dynamic programming (ADP) algorithm has strict restrictions on the system, which must satisfy the condition that the equilibrium point is zero and $f\left(0\right)=0$. However, in practice, these limits are difficult to achieve. To overcome this problem, we design a new cost function that successfully removes this limitation. At the same time, in order to improve the motion accuracy and control effect, the effects of joint friction torque and interaction forces between the LLER and the user on the system dynamics are considered. Aiming at the difficulty of solving the Hamilton–Jacobi–Bellman (HJB) equation, a critic neural network learning framework is designed to approximate the optimal cost function, and auxiliary terms are introduced to eliminate the requirement of initial stability control. Throughout the entire learning process, the update of the controller is driven by an event-triggered mechanism, which significantly reduces the computational burden on the robotic system. Finally, the effectiveness of the proposed algorithm is verified through simulation experiments.