The Improved Simulated Annealing Genetic Algorithm For Robot Dynamics Parameter Identification

In order to meet the requirements of the robot's working speed and accuracy,the control method based on the robot dynamics model is widely used.How to use the dynamic parameter identification method to optimize the robot dynamics model is a hot issue in robot dynamics research.In terms of dynamic parameter identification algorithms,compared with least square method,maximum likelihood method,neural network algorithm,etc.,genetic algorithms have certain advantages in robustness,adaptability and parallelism,and there are broad prospects for solving the problem of parameter identification for robots and other nonlinear systems.The paper proposes an improved simulated annealing genetic algorithm to identify the dynamic parameters of the robot.Firstly,the paper describes the main methods of dynamic modeling of industrial robots,and introduces the important parameters of the dynamic characteristics required for dynamic modeling.Secondly,according to the robot force sensor signal,the robot dynamics parameters are identified by the Newtonian Euler dynamic recurrence equation,and the identification steps are given.Then,the improvement,optimization and fusion based on basic genetic algorithm and basic simulated annealing algorithm,the improved simulated annealing genetic algorithm is proposed.The improved simulated annealing genetic algorithm improves by using floating point encoding and inter-cell generation of initial populations,genetic manipulation adopts group selection,adaptive crossover that mixs arithmetic crossover with heuristic crossover and adaptive variation.After genetic operation of the improved genetic algorithm,the new individual offspring population mechanism combines the improved simulated annealing algorithm cooling schedule Metropolis criteria,which improves the efficiency and accuracy of the improved algorithm.Finally,PUMA560 is used as the simulation object.The improved simulated annealing genetic algorithm is used to identify the dynamic parameters after selecting the optimal trajectory of each joint.Compared with the least-squares method,conventional genetic algorithm and single genetic algorithm improves by Floating-Point Encoding,Inter-Cell Generation,and Improved Genetic Operation,the simulation data shows that the improved simulated annealing genetic algorithm has great advantages in terms of optimization time,optimization accuracy and themaximum error of the identification parameters on solving the robot parameter identification problem compared with the first three algorithms.