Abstract: Hyperbolic tangent function is used instead of sigmoid funtion as the action function for the artificial neural networks，which used to descript environment.This makes the ANN（Artificial Neural Networks） ideal for optimization arithmetic.PSO（Particle Swarm Optimization） has the advantage of fast convergence speed and few parameters to adjust，but premature convergence often occurs during optimization.SA（Simulated Annealing），intercross operator and aberrance operator are combined to improve PSO’s performance，a new IPSO（Improved Particle Swarm Optimization） is formed to solve AGV（Automated Guided Vehicles）’s global path planning problem.Simulation results demonstrate IPSO’s ability of seeking the global result is strong，convergence speed is fast than PSO.IPSO can plan a ideal path for AGV.

Key words: Automated Guided Vehicles（AGV）, global path planning, Improved Particle Swarm Optimization（PSO）, Artificial Neural Networks（ANN）, action function

摘要： 分析了用人工神经网络模型描述环境时，采用Sigmoid函数作为神经网络作用函数的不足之处，提出采用双曲正切函数作为神经网络的作用函数，使网络更有利于路径优化算法的寻优计算。粒子群优化（Particle Swarm Optimization，PSO）算法具有收敛速度快，需要调节的参数少等优点，但优化过程中容易发生“早熟”收敛，使优化陷入局部极小值。通过引入模拟退火算法、“交叉算子”和“变异算子”，提出了一种新的改进粒子群优化算法（Improved Particle Swarm Optimization，IPSO）来解决AGV全局路径规划问题。仿真结果表明，IPSO具有很强的全局寻优能力，并且收敛速度比PSO快，能够为AGV规划出理想的路径。

关键词: 自动牵引车, 全局路径规划, 改进粒子群优化, 人工神经网络, 作用函数