Optimal online generation of obstacle avoidance trajectory running on a low speed embedded CPU for vehicles

Considering obstacle avoidance for mobile robots, it is effective to generate optimal trajectory dynamically in terms of safety and efficiency. Since it usually requires iterative calculation to get an optimal trajectory, however, high computational cost becomes a problem for practical application, especially for low performance computers like industrial embedded CPUs. In this paper, we propose a low cost trajectory optimization method which is applicable to embedded CPUs. This method suppresses the computational cost through introduction of a low order polynomial function and proper boundary conditions to generate continuous trajectory. In order to see the feasibility, we applied proposed method to a wheeled robot which is equipped with an embedded CPU and a laser range finder (LRF) to show the performance of obstacle avoidance experiments.