Fast trajectory planning of reactive operation considering manipulability for a humanoid

For humanoid robots, there exist a class of challenging tasks, e.g. playing table tennis, which are non-repeatedly, triggered randomly by sensor events, and compelling the robots to produce reactive operations in very short time. Additionally, some key poses of the robot are partially constrained to accomplish the operations, and are usually adjusted during the operation execution. In this paper, efforts are made to generate trajectory for these operations. Firstly, an extended manipulability index is proposed to optimize the key poses so that the humanoid robot has the manipulability considering joint limits at the key poses, which makes it convenient to adjust the trajectory. Secondly, a fast approach to generating the whole trajectory consisting of all the key poses and the strategies to realize the trajectory adjustment when specifications change are proposed. Finally, the ZMP (Zero Moment Point) displacement excited by the reactive operations is also given to decide whether the generated trajectory is feasible. The effectiveness of the method is demonstrated through simulations.