Robotic organism targets regional coverage capture path planning for marine aquafarm based on value iteration network

Abstract

The application of robots in the field of Marine aquatic organisms capture can greatly improve the fishing efficiency and economy of seafood. The purpose for path planning of underwater vehicle manipulator system (UVMS) is to capture the most organisms in the shortest time with no obstacle collisions. This paper has proposed a novel cost modelling algorithm for the collision avoidance, manipulation and vehicle movement during the target capture process. On the underwater unknown and blurred environment, a novel double value iteration network with long short-term memory (LSTM) network has been proposed on the basis of value iteration networks. The proposed improved double value iteration network can predict the blurred and unknown environment, plan capture strategy with the calculated collision risk and movement cost. In the value iteration module, the use of LSTM network can improve the stability of training and the performance of the model. Moreover, a novel cooperative path planning model for task allocation has been proposed to improve capture efficiency. Simulations on robot and multiple robots capture planning have verified the network performance.