一种单计算参数的自学习路径规划算法
|
摘要
针对当前机器人路径规划算法存在计算参数多的问题,提出一种单计算参的自学习蚁群算法。该算法使用一种改进的栅格法完成环境建模,种群中个体使用8-geometry行进规则,整个种群的寻优过程使用了自学习和多目标搜索策略。其特点在于整个算法只需进行一个计算参数设置。蚂蚁个体可使用1、■、2、■、■步长行进,一次搜索可以发现多条可行路径,提高了算法计算效率。仿真实验表明,在复杂的工作空间,该算法可以迅速规划出一条安全避碰的最优路径,效率优于已存在算法。
The existing robot path planning( RPP) algorithms have the problems that the parameters are complexity. To solve this problem, this paper proposes a self-learning ACO(SlACO) algorithm for robot path planning. In SlACO, an improved grid map(IGM)method is used for modeling the working space and the 8-geometry is used as the moving rule of ant individuals. The strategy of multi-objective search is used for the whole ant colony. The SlACO has the feature that the whole algorithm only need set one computing parameter. Moreover, the ant individuals can move with step 1, ■, 2, ■ and ■. By the strategy of machine learning and multi-objective search, the SlACO algorithm can find more than one feasible paths with a move from starting position to the ending position. Simulation results indicate that the SlACO algorithm can rapid plan a smooth even in the complicated work-ing space and its efficiency is better than existing RPP algorithms.
The existing robot path planning( RPP) algorithms have the problems that the parameters are complexity. To solve this problem, this paper proposes a self-learning ACO(SlACO) algorithm for robot path planning. In SlACO, an improved grid map(IGM)method is used for modeling the working space and the 8-geometry is used as the moving rule of ant individuals. The strategy of multi-objective search is used for the whole ant colony. The SlACO has the feature that the whole algorithm only need set one computing parameter. Moreover, the ant individuals can move with step 1, ■, 2, ■ and ■. By the strategy of machine learning and multi-objective search, the SlACO algorithm can find more than one feasible paths with a move from starting position to the ending position. Simulation results indicate that the SlACO algorithm can rapid plan a smooth even in the complicated work-ing space and its efficiency is better than existing RPP algorithms.
引文
[1]陈明建,林伟,曾碧.基于滚动窗口的机器人自主构图路径规划[J].计算机工程,2017,43(2):286-292.
[2]殷路,尹怡欣.基于动态人工势场法的路径规划仿真研究[J].系统仿真学报,2009,21(11):3325-3341.
[3]刘毅,车进,朱小波,等.空地机器人协同导航方法与实验研究[J].电子技术应用,2018,44(10):144-148.
[4]尹新城,胡勇,牛会敏.未知环境中机器人避障路径规划研究[J].科学技术与工程,2016,16(33):221-226.
[5]唐焱,肖蓬勃,李发琴,等.避障最优路径系统研究[J].电子技术应用,2017,43(11):128-135.
[6]ERGEZER H,LEBLEBICIOGLU K.Path planning for UAVs for maximum information collection[J].IEEE Transac tions on Aerospace and Electronic Systems,2013,49(1):502-520.
[7]PAMOSOAJI A K,HONG K S.A Path-planning algo rithm using vector potential functions in triangular regions[J].IEEE Transactions on Systems,Man,and Cybernetics:Systems,2013,43(4):832-842.
[8]WU N Q,ZHOU M C.Shortest routing of bidirectional automated guided vehicles avoiding deadlock and blocking[J].IEEE/ASME Transactions on Mechatronics,2007,12(2):63-72.
[9]田延飞,黄立文,李爽.激励机制改进蚁群优化算法用于全局路径规划[J].科学技术与工程,2017,17(20):282-287.
[10]徐晓晴,朱庆保.基于蛙跳算法的新型机器人路径规划算法[J].小型微型计算机系统,2014,35(7):1631-1635.
[11]朱庆保.复杂环境下的机器人路径规划蚂蚁算法[J].自动化学报,2006,32(4):586-593.
[2]殷路,尹怡欣.基于动态人工势场法的路径规划仿真研究[J].系统仿真学报,2009,21(11):3325-3341.
[3]刘毅,车进,朱小波,等.空地机器人协同导航方法与实验研究[J].电子技术应用,2018,44(10):144-148.
[4]尹新城,胡勇,牛会敏.未知环境中机器人避障路径规划研究[J].科学技术与工程,2016,16(33):221-226.
[5]唐焱,肖蓬勃,李发琴,等.避障最优路径系统研究[J].电子技术应用,2017,43(11):128-135.
[6]ERGEZER H,LEBLEBICIOGLU K.Path planning for UAVs for maximum information collection[J].IEEE Transac tions on Aerospace and Electronic Systems,2013,49(1):502-520.
[7]PAMOSOAJI A K,HONG K S.A Path-planning algo rithm using vector potential functions in triangular regions[J].IEEE Transactions on Systems,Man,and Cybernetics:Systems,2013,43(4):832-842.
[8]WU N Q,ZHOU M C.Shortest routing of bidirectional automated guided vehicles avoiding deadlock and blocking[J].IEEE/ASME Transactions on Mechatronics,2007,12(2):63-72.
[9]田延飞,黄立文,李爽.激励机制改进蚁群优化算法用于全局路径规划[J].科学技术与工程,2017,17(20):282-287.
[10]徐晓晴,朱庆保.基于蛙跳算法的新型机器人路径规划算法[J].小型微型计算机系统,2014,35(7):1631-1635.
[11]朱庆保.复杂环境下的机器人路径规划蚂蚁算法[J].自动化学报,2006,32(4):586-593.