| Redundant robot can optimize performance, such as avoiding obstacles,avoiding singularity and avoiding joint range overrun with its redundant performance,to largely improved flexibility of the robot.Gradient projection method which is the most basic method of redundant robotto solve kinematics problem, can optimize the performance of redundant robot. Butgradient projection method is a local optimization algorithm, so it is easy to fail intolocal optimum in the process of optimizing algorithm, at the same time it can’tguarantee that redundant robot finally get the optimal solution. In order to preventthis phenomenon, this paper adopts the evolution algorithm with global optimalspecialty to optimize performance of the redundant robot. Firstly the paperintroduces the basic principle of differential evolution algorithm, analyzes the non-dominated sorting genetic algorithm and the strength non-inferior evolutionaryalgorithm, also studies the optimization operator and optimization process of the twoalgorithms; Then in order to improve the algorithm convergence and keep thediversity of solutions, this paper improves the two algorithms. The differenceoperator, mutation and crossover operation of difference arithmetic is brought intothe two algorithms, so the non-dominated sorting genetic algorithm differentialevolution algorithm and the strength non-inferior differential evolution algorithm aredeveloped.Finally, this paper takes three degree-of-freedom redundant robot as researchobject, according to the robot end trajectory planning, validates the above algorithms.The simulation results show that in the aspect of optimization multi-objective toredundant robot, the genetic algorithm based on difference evolutionary algorithm can realize the global optimal and the probability of getting into local optimum is low,the optimization effects is more significant compared to other multi-objectiveoptimization algorithm and gradient projection method. |
PDF Full Text Request