Research on Path Planning Algorithm for Feeding Robots Based on Deep Reinforcement Learning

The current aquaculture industry is facing dual challenges of dynamic environment and multitasking demands. This article proposes a targeted dynamic task priority adjustment mechanism based on deep reinforcement learning algorithm - MTP-DDPG, to solve the multi-objective path planning problem of multiple feed buckets. This method dynamically adjusts task priority by introducing factors such as task urgency, task weight, and environmental changes, enabling the feeding robot to optimize path selection in complex environments, thereby improving the training speed and efficiency of the algorithm on multi-objective tasks. Conduct simulation experiments on feeding robots in dynamic scenarios in ROS. The simulation results show that the MTP-DDPG algorithm can effectively complete feeding tasks in complex and multi-objective scenarios.