塔里木大学学报2026，Vol.38Issue(3)：88-95,8.DOI:10.3969∕j.issn.1009-0568.2026.03.009
基于A2C算法的万寿菊采摘机器人智能控制方法
Intelligent control method for marigold harvesting robot based on advantage actor-critic(A2C)algorithm
摘要
Abstract
With the continuous expansion of marigold cultivation in the Southern Xinjiang region of China,the traditional manual harvesting model,plagued by inefficiency and high labor intensity,has struggled to meet the demands of industrial-scale development.To enhance the level of mechanization and intelligence in the harvesting process,an autonomous control system based on the advantage actor-critic(A2C)reinforcement learning algorithm was developed.Centered on this system,a machine vision-integrated pneumatic marigold harvesting robot was designed.The robot uses a machine vision module to accurately locate flower positions and relies on the reinforcement learning algorithm to achieve autonomous motion decision-making and picking action execution.System performance was evaluated by constructing a laboratory test platform that simulated marigold field growth conditions.Test results show that the flower recognition success rate reached 90.5%,and the picking success rate reached 92.5%,demonstrating the robot's ability to stably complete the entire workflow from visual perception to motion control and picking execution.This paper confirms the feasibility of intelligent control methods in marigold harvesting scenarios and offers important technical support for improving efficiency and reducing labor costs in the marigold industry in Southern Xinjiang.关键词
万寿菊采摘机器人/强化学习/A2C算法/自主采摘Key words
marigold harvesting robot/reinforcement learning/advantage actor-critic algorithm/autonomous harvesting分类
信息技术与安全科学引用本文复制引用
杨凯琳,陈立平..基于A2C算法的万寿菊采摘机器人智能控制方法[J].塔里木大学学报,2026,38(3):88-95,8.基金项目
南疆重点产业创新发展支撑项目(2023AB040) （2023AB040）
