基于云-边协同强化学习的多桁架机器人避碰控制方法  

Cloud-Edge-Based Collaborative Reinforcement Learning for Collision Avoidance Control of Multiple Truss Type Robots

作　　者：吕洋 林志诚 Lv Yang

机构地区：[1]东南大学自动化学院,江苏南京210096

出　　处：《工业控制计算机》2025年第1期1-3,共3页Industrial Control Computer

摘　　要：桁架机器人已在大型、重载搬运与作业场景中发挥了重要作用。当前同导轨上多桁架机器人在协同作业过程中的连续避碰控制的需求广泛且迫切。提出了一种基于云-边协同强化学习的多桁架机器人连续避碰控制方法。采用了云-边计算架构下的深度确定性策略梯度算法,通过云端在线学习、边缘离线推理的手段实现了两两相邻桁架机器人的平滑、连续的避碰控制。实验结果表明,相较于传统算法,该方法给出的运动轨迹更为平滑且具备更快地作业速度,很好地满足了多桁架设备协同作业中的鲁棒性和实时性需求。In this paper,a continuous collision avoidance control method for multiple truss type robots based on cloud-edge collaborative reinforcement learning is proposed.The deep deterministic policy gradient algorithm under the cloud-edge computing architecture is adopted,and the smooth and continuous collision avoidance control of two neighboring truss type robots is realized by means of online learning in the cloud and offline inference at the edge.The results show that compared with the traditional algorithm,the motion trajectory given by this method is smoother,which well meets the robustness and real-time demand in the cooperative operation of multi-truss equipment.

关 键 词：云-边协同 强化学习 多桁架机器人 避碰控制 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]