Adaptive Control of Fully-Actuated Cable-Driven Parallel Robots for Mars Rover Landing Simulation

In this paper, an adaptive nonsingular terminal sliding mode control (ANTSMC) is proposed for cable-driven parallel robots (CDPRs) used in the Mars rover landing simulation. Accurate position accuracy of the CDPRs is crucial for simulating the Mars rover landing environment. To solve the above issue, an ANTSMC is innovatively proposed to handle the uncertainties, where the adaptive parameters are obtained by the deep reinforcement learning (DRL)-based parameter adaption model. The DRL-based parameter adaption model can dynamically identify the optimal parameters, improving the robustness and intelligence. Compared to traditional methods, the parameter adaption model avoids overly relying on the exact system models with wider potential for applications. Validation simulations are conducted, and the results demonstrate that the ANTSMC significantly enhances the robustness and improves tracking accuracy. Compared to the classical augmented proportion-derivative, the proposed method reduces the tracking error by about 50 % in statistical terms.