2018 年 9 巻 2 号 p. 39-47
A novel trajectory generator for obstacle avoidance is proposed and evaluated through numerical simulations with common iterative methods. Since the proposed method generates quasi-optimal trajectories using model predictive control (MPC) theory with a predetermined upper bound on computational cost, it makes it easy to guarantee real-time feasibility for autonomous driving and/or driving support systems. Numerical round-robin simulations are conducted for both the proposed method and a comparative method, after which we evaluate the results through statistical analysis and individually analyze several characteristic results. Taken together, the results show that the proposed method generates trajectories that are statistically equivalent to those generated by the comparative method, while guaranteeing that the upper bound of the computational cost is predetermined.