Yazi Zhang, Guoying Han, and Bin Zhou
CMA-ES, parameter clustering, chaotic mapping, trajectory generation, robot arm
The humanoid robot arm can cooperate with other institutions to complete a series of anthropomorphic behaviours, such as grasping, carrying, throwing, etc., and can also replace human beings to perform specific tasks in complex environments. Aiming at the trajectory optimisation of the arm of the acquisition robot, D-H modelling method and inverse transformation method are used to conduct forward and inverse kinematics modelling of the right arm of the robot. The problem of slow convergence of the multi-dimensional parameter optimisation of the robot is solved by the optimisation method of parameter curve similarity clustering. Finally, chaos mapping and Levi flight disturbance are introduced. The goal is to solve the problem that covariance adaptive evolution is prone to local optimisation. The results showed that the overall accuracy of the proposed trajectory optimisation algorithm was improved by 32%, and the values of the four objective functions were reduced by 75.09%, 92.36%, 47.06%, and 10%. This indicated that the global optimisation ability was significantly improved, and the optimisation trajectory was smoother. The average maximum error at the key joints was 0.48, with high precision. The covariance adaptive evolutionary strategy optimisation algorithm proposed in this study has better arm trajectory generation performance and can achieve the effect of completely tracking the trajectory with fewer iterations.
Important Links:
Go Back
