Hierarchical Trajectory Optimization for Humanoid Robot Jumping Motion.

In order to make the robot have high dynamic motion ability, this paper uses the trajectory optimization method to plan the jumping motion trajectory of the humanoid robot quickly and effectively, and carries out the jumping motion experiment on the HIT-HU humanoid robot platform. In order to improve the solving speed of trajectory optimization problem, simplified centroid dynamics model and single rigid body model are established. Differential dynamic programming (DDP) based on the combination of the centroid dynamic model and the whole body joint kinematic model is used to solve the hopping trajectory optimization problem with high accuracy, and the external penalty function method is used to deal with the relevant constraints. In order to overcome the problem that the solver is sensitive to the initial point selection, a hierarchical trajectory optimization framework is proposed. The framework uses the direct collocation method based on the single rigid body model. The trajectory of the simplified model is used as the initial trajectory guess, and the differential dynamic programming method is introduced to solve the optimal trajectory of the whole joint. The differential dynamic programming algorithm using the direct method of warm start is about twice as fast as the method of manually given initial trajectory. The simulation and experimental verification of the vertical jump trajectories generated by the algorithm on the humanoid robot HIT-HU show that the jump trajectories generated by the algorithm can be effectively transformed into the motion trajectories of humanoid robots, and can be effectively executed on the physical prototype. This further verifies the effectiveness of the algorithm and plays an important role in practical application.