Abstract:

To solve the problem of component grasping path planning, a hybrid optimization algorithm based on ant colony algorithm and tabu search algorithm was proposed to minimize the time. Firstly, the problem of grasping components based on machine vision was defined as Traveling Salesman Problem (TSP) with precedence constraint. Secondly, comprehensive effects of the sizes of components and the grasping and placement process on path planning were analyzed, and the path selection probability and tabu region were improved adaptively. Thirdly, on the one hand, 2-opt local optimization, pheromone punishment and reward mechanism were introduced to improve the search ability of ants; on the other hand, pheromone evaporation factor was improved adaptively to increase the adaptability of ants. Finally, for the basic algorithm and the improved hybrid optimization algorithm, the performance index and grasping time were compared and analyzed by simulation experiment and platform experiment. The experimental results show that, compared with Ant Colony Optimization (ACO) algorithm and Tabu Search (TS) algorithm, the average iteration times of the proposed hybrid optimization algorithm is reduced by about 50%, and other performances are superior to the other algorithms. The results of platform test also show that the hybrid optimization algorithm is superior to random results and basic algorithm, and it can realize component grasping task quickly.

Key words: machine vision, Traveling Salesman Problem (TSP) with precedence constraint, optimal path, ant colony algorithm, hybrid optimization algorithm

摘要：

针对元件的抓取路径规划问题，提出一种以最小化时间为目的，结合蚁群算法和禁忌搜索算法的混合优化算法。首先，将基于机器视觉抓取元件的问题确定为有约束的旅行商问题（TSP）；然后，分析了元件大小和抓取放置过程对于路径规划的综合影响，对路径选择概率和禁忌域进行了适应性改进；其次，一方面引入了2-opt局部优化以及信息素惩罚、奖励机制以改善蚂蚁的搜索能力，另一方面对信息挥发因子作适应性改进以提高蚂蚁的自适应能力；最后，针对基本算法和改进的混合优化算法，仿真实验和平台实验分别进行了性能指标和抓取时间的对比分析。实验结果表明，仿真环境下，与蚁群优化（ACO）算法和禁忌搜索（TS）算法相比，混合优化算法的平均迭代次数降低了约50%，且其他性能较为优越，平台测试的抓取用时测试结果也说明了混合优化算法较随机结果和基本算法的优越性，可以快速完成元件抓取任务。

关键词: 机器视觉, 有约束的TSP, 最优路径, 蚁群算法, 混合优化算法