Innovative Approaches to Data-Driven Inverse Optimal Control in Continuous-Time Linear Systems

This paper aims at addressing the challenges in inverse optimal control (IOC) for continuous-time (CT) linear systems. To this end, a novel computation framework is employed to design reward matrices and generate optimal control laws for tracking target system trajectories, thereby enhancing computational efficiency, alleviating overstrict initialization requirements and achieving better tracking performance. Firstly, a model-based IOC algorithm is proposed, which utilizes the gradient descent method to directly tune the reward matrix. Subsequently, by fully utilizing target data and employing the adaptive dynamic programming (ADP) technique, a model-free IOC algorithm is developed, where only one integral equation needs to be solved. The proposed algorithms are proven to converge to the optimal solution, and the stability of the generated control laws is guaranteed. Finally, two simulation results are presented to demonstrate the effectiveness of the proposed methods.