Dynamic Task Planning for Multi-Arm Apple-Harvesting Robots Using LSTM-PPO Reinforcement Learning Algorithm

Abstract

This paper presents a dynamic task planning approach for multi-arm apple-picking robots based on a deep reinforcement learning (DRL) framework incorporating Long Short-Term Memory (LSTM) networks and Proximal Policy Optimization (PPO). In the context of rising labor costs and labor shortages in agriculture, automated apple harvesting is becoming increasingly important. The proposed algorithm addresses key challenges such as efficient task coordination, optimal picking sequences, and real-time decision-making in complex, dynamic orchard environments. The system’s performance is validated through simulations in both static and dynamic environments, with the algorithm demonstrating significant improvements in task completion time and robot efficiency compared to existing strategies. The results show that the LSTM-PPO approach outperforms other methods, offering enhanced adaptability, fault tolerance, and task execution efficiency, particularly under changing and unpredictable conditions. This research lays the foundation for the development of more efficient, adaptable robotic systems in agricultural applications.