A BP-GA Approach to Solve the Inverse Kinematics of Virtual Human's Upper Limb Kinematic Chain

摘要

The human's Upper Limb Kinematic Chain (ULKC) is the most frequently used and most complicated kinematic chain. It is difficult to compute its Inverse Kinematic (IK) solution quickly and accurately by using neural network or Genetic Algorithm (GA) due to ULKC's high degree of freedom. Combining BP neural network and GA, we proposed a method to solve its IK problem. Firstly, the joint-units of ULKC and its mathematical model were constructed based on D-H method. Then the BP neural network produced a local optimal solution, which could be served as an individual of GA's initial population. We could determine the searching domain by its optimal solution. Finally, the high accuracy solution was obtained by using the adaptive GA. The experimental results showed that the proposed approach could obtain the high accuracy solution efficiently with the BP neural network high-speed and GA high-accuracy.