Research On Path Planning Of Humanoid Robot Based On The Genetic Algorithm

Humanoid robot is a kind of mobile robot with humanoid structure. Becausehumanoid robot has more freedom degree, the complexity of the behavior control andunique system mechanical structure, the movement process needs a stable footstepcontrol method to keep the whole balance. With more freedom rotating steering gear arecontrolled to achieve movement and keep the center of gravity balance, which needsenergy consumption, from the independent movement of the humanoid robot carryingenergy to see, it is limited after all, to get the energy saving and maximize using, at thesame time, it also includes in path planning that the distance, path smoothness, andother target evaluation function are applied minimization principle. So this is a typicalmulti-objective optimization problem in path planning.According to the development of multi-objective optimization problem and theresearch of its algorithm, it has provided convenience to solve the humanoid robot pathplanning in the multi-objective optimization problem. Multi-objective optimizationgenetic algorithm(MOGA) is proposed to have many advantages solve the problem ofthe optimization of the many targets, with going forward hand in hand manyimprovement, the MOGA is now more mature, fast the non-dominate sorting geneticalgorithm(NSGA) can very well evaluate multiple targets and be selected to solvemulti-objective optimization problem in path planning. And test functions are conducteda series of performance testing, it is proved that this algorithm has fast convergence rateand good robustness, so it is used to solve multi-objective optimization problem in pathplanning.Due to the grid method is simple and effective, and it is used to realize theenvironment modeling, a path can be generated with a series of random sequence gridnumber, it contains a starting point and goal point, which is a randomized path.According to the relationship of path sequence point and obstacles, random path can bedivided into feasible path and infeasible path, but feasible path is better than infeasiblepath, which function value is smaller. According to the humanoid robot path planningrequirements, in the paper crossover operator and mutation operator are improved, andthe optimization operator is used, these operators can increase population diversity andimprove the convergence, according to initializing random produced population, which is sorted non-dominantly, the population can get different levels of optimizationindividuals, and crowded degree comparison operator is used to find the approximateoptimal individual group, which are kept as elite group, and elite group gets geneticvariation and iterative optimization generation by generation, Pareto optimal solutionsare close to the real solutions,this individual is sure to a low level in non-dominatesorting level. By adjusting the three parameters in NSGA-II: population scale, genelength and number of generation, according to different final optimization results andthe comparison of Pareto distribution, so the optimum parameters are determined, eachrun of the algorithm can produce many Pareto-optimal solutions, which provides aneffective tool for measuring the performance of different objective functions,simulations show that the method in this paper is feasible and efficient.