基于A2C算法的万寿菊采摘机器人智能控制方法OACHSSCD

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.