基于XCS和LS-SVM的ALV在狭隘环境中的避碰规划

Abstract

动态环境下的局部避碰是自主地面车的一项基本功能，在自主导航中占有重要作用。由于遗传算法具有早熟收敛、局部最优解和占据较大的存储空间等缺陷，为了提高自主地面车的避碰能力，本文提出了一种基于学习分类器的自主地面车在狭隘环境中路径避碰规划方法，设计改进了特殊的遗传算子。不同环境的仿真实验结果表明LS-SVM和学习分类器结合用于自主地面车的路径规划是收敛的，提高了ALV在狭隘环境中快速发现安全路径的能力。 Local obstacle avoidance in dynamic narrow environments, as a principal ability for ALV, plays an important role in autonomous navigation. Due to premature convergence, local optimal solution, accounting for a larger storage space and other shortcomings of genetic algorithms still exist. In order to improve the ability of obstacle avoidance for ALV, this paper presents a path obstacle avoidance planning method based on LCS for ALV in narrow environments, designs and improves special Genetic Operators. Different environments of simulation results showed that the combination of LS-SVM and learning classifier for ALV path obstacle avoidance planning was convergent, increasing ALV’s ability to quickly find safe paths in narrow environments.