非完整运动规划的粒子群优化算法

讨论了粒子群优化和非完整运动规划问题。首先对粒子群优化算法的性能进行了分析，发现当搜索空间的维数较高时，粒子群将收敛到子空间的一个局部最优点，而该点并不是整个搜索空间的局部最优点。通过引入变异算子，则可以改进粒子群优化算法的性能。在此基础上，提出了一种求解非完整运动规划问题的带变异算子的粒子群优化算法。仿真结果表明，对于30维的球形函数，无变异操作的粒子群优化算法不能得到问题的最优解，而最优解可以非常容易地由带变异操作的粒子群优化算法得到。对独轮移动机器人非完整运动规划问题进行了仿真。结果表明，粒子群优化算法比牛顿法更有效。

Particle swarm optimization for nonholonomic motion planning

A particle swarm optimization (PSO) and the nonholonomic motion planning are discussed.The performance of the PSO is investigated first.It has been found that when the dimension of search space is higher,the particles will be converged at a local optimum point in a subspace,but not an optimum one in the whole search space.The mutation can be introduced to improve the performance of PSO.A PSO algorithm with mutation is proposed to resolve the nonholonomic motion planning.The simulation results show:For a 30 dimension sphere function, the optimum point can not be found by PSO without mutation,but can be obtained easily by PSO with mutation.The simulations are also done for the nonholonomic motion planning of a unicycle mobile robot.The results show that the PSO algorithm proposed by this paper is more effective than Newton algorithm.