Data-driven Predictive Control for Safe Motion Planning

Controlling a constrained dynamic system in an environment with multiple obstacles is important yet challenging. Many existing methods are either heuristic (e.g. A^* algorithm) or model-based (e.g. optimal control). In contrast to these methods, this paper addresses scenarios where the mathematical model of the dynamic system is unknown, relying solely on input–output data and environmental information. We propose a new data-driven framework to achieve safe path planning and efficient tracking control by integrating sample-based methods with more recent data-enabled predictive control. In the offline phase, we develop a safe path planning algorithm to generate a sequence of convex safe sets from the initial point to the target set. This is achieved by leveraging a sample-based planning algorithm and solving bi-linear optimization problems. The resulting adjacent safe sets have a nonempty intersection, and the distance between each safe set and any obstacle exceeds the required safe distance. In the online phase, we develop an efficient data-enabled predictive tracking control algorithm with the core of safe set contraction constraints to sequentially track the safe sets. The proposed algorithm transforms the nonconvex obstacle avoidance control problem into a convex optimization problem, which can be solved efficiently. We demonstrate that the proposed framework is safe, efficient, and scalable through quadcopter simulations, comparison simulations, and unmanned vehicle experiments.