融合神经网络及麻雀算法的机器人避障研究

Research on Robot Obstacle Avoidance Based on Neural Network and Sparrow Algorithm

针对麻雀搜索算法(SSA)在机器人避障研究中,存在提早收敛于局部最优难以跳出、初始种群分布不够广泛、平衡能力差等问题对其进行改进;通过三层神经网络对规划环境进行栅格化建模;引入Halton序列得到初代种群分布,得到分布更广、更遍历的个体位置,提升后期寻优速度和效率;使用布朗运动优化麻雀位置更新的步长调节,帮助算法脱离局部优解,同时平衡全局切换局部的搜索节奏;利用clothoid曲线法平滑路径,得到满足机器人机械性能的路径;经6个标准函数验证和Wilcoxon检验P值对比可知,改进后的算法相较于SSA和CSSA算法各项指标得到明显优化,且具有和SSA同一水平的时间复杂度。通过地图仿真得到平滑后的机器人避障路径。

Aiming at the problems of a sparrow search algorithm(SSA)in a robot obstacle avoidance research,such as early convergence to the local optimal difficult to jump out,insufficient distribution of the initial population,poor balance ability,in order to improve the questions,a three-layer neural network was used to model the planning environment raster.The Halton sequence was introduced to obtain the initial population distribution,obtain the individual locations with wider distribution and more ergodic distribution,and improve the speed and efficiency of later optimization.The Brownian motion was used to optimize the step size adjustment of the sparrow position update,which helped the algorithm out of local optimal solution and balanced the local search rhythm of global switch.The method of the clothoid curve was used to smooth the path,and obtain the satisfied path with the mechanical properties of the robot.Compared with the algorithms of the SSA and chaos sparrow search algorithm(CSSA),the improved algorithm is significantly optimized in the P-values of the Wilcoxon test and six standard functions,and has the same level of time complexity as the SSA.The smooth obstacle avoidance path of the robot is obtained by the map simulation.