全身動力学モデル予測制御における高速な自己衝突回避

抄録

This paper proposes self-collision avoidance for whole-body model predictive control (WB-MPC). Since WB-MPC requires solving a large-scale optimization problem, the gradients of the dynamics and cost functions must be computed quickly. To compute the gradient of the self-collision detection quickly, we create a collision detector using a deep neural network. The DNN-used collision detector estimates the minimum distance between all body frames based on the whole body joint angles, converting to a cost. As a result, it enables WB-MPC to plan long-term self-collision-aware motion by using its gradient calculated at high speed. The proposed method was implemented for a physics simulation using a dual-arm robot and outperformed the previous method with inverse kinematics in terms of long-term planning with self-collision avoidance.