Trajectory Generation Based on Model Predictive Control with Obstacle Avoidance between Prediction Time Steps

In this paper, model predictive control (MPC) based trajectory generation for a unicycle with obstacle avoidance is addressed. Since MPC based methods can only consider obstacle avoidance at discrete time steps, a collision may occur in intervals between prediction time steps. In order to prevent this problem, we propose two methods which guarantee the obstacle avoidance not only at prediction time steps but also in intervals between them. The first method reduces the maximum velocity near obstacles. The second one constrains transition of collision avoidance constraints to exclude possibilities of collisions between time steps. Since both methods do not need iteration of optimization, it is expected that the computation time is not significantly increased compared with existing methods. Numerical examples and experiments show the effectiveness of the proposed methods.