Autonomous driving has seen increased interest nowadays. The coordination of multi-vehicle is a key issue in autonomous driving. While many people are trying to tackle the problem using the knowledge of A.I. (machine learning and deep learning), this project discusses the problem from a different angle–control theory. In this project, I propose a controller for the multi-agent system which can achieve the following objectives: (1) collision avoidance, which means no vehicle will bump into each other during the process (2) finite-stability, a stronger version of Lyapunov stability, which means the vehicles can arrive at their destinations in finite time. Simulation results are given to prove the efficacy of the controller.