Motion learning for redundant manipulator with structured intelligence
Naoyuki Kubota, Takemasa Arakawa, Toshio Fukuda · 2002
This paper deals with trajectory planning and motion learning for a redundant manipulator. We have proposed a hierarchical trajectory planning method by a virus-evolutionary genetic algorithm. Furthermore, we have applied a neural network for the motion learning of trajectories generated by the hierarchical trajectory planning method. This paper proposes a primitive motion planning method by using outputs of the learned neural network. The simulation results show that the primitive motion planning method can reduce computational cost and quickly obtain collision-free trajectories of a redundant manipulator.