基于遗传神经网络的车辆导航路径规划

研究使用混合 GA-BP 神经网络算法来解决交通路径规划中的非线性问题。反向传播(Back-Propagation, BP)神经网络虽然能够很好地解决非线性问题,但它存在着容易陷入局部极小的不足,而遗传算法(Genetic Algorithm, GA)具有很强的宏观搜索能力和良好的全局优化性能,可以弥补BP的不足。用A*算法快速粗算出的几条可选路径作为 GA 的初始种群,然后用混合的 GA-BP 神经网络算法进行路径规划精算。仿真结果显示混合GA-BP神经网络算法在寻找路径规划的全局最优解上具有一定的优势。

Path Planning Approach to Vehicle Navigation Based on Genetic Neural Network

In this paper, we solved nonlinear problems in the traffic path planning with the hybrid GA-BP neural network algorithm. Although Back-Propagation neural network (BP) is able to solve nonlinear problems properly, it is tend to fall into the deficiency of local minimum. In the meanwhile, genetic algorithm (GA) is good at macro-searching and performs well at global optimization, which can make up for the deficiencies of BP. In this paper, using the A^* algorithm, we rough calculated several alternative paths quickly, which serve as the initial population of the GA. Then we conducted path planning precisely with the mixed GA-BP neural network algorithm. The simulation results showed that the hybrid GA-BP neural network algorithm has some advantages in the global optimal solution for path planning.