TARGET TRACKING SYSTEMS AND METHODS FOR UUV

摘要

As the marine environment becomes more complex, the good target trackingsystem and methodis essential for unmanned underwater vehicle (UUV). Firstly, the UUV trackingsystem isproposed, which includes Control System, Sensor System, Propulsion System,CommunicationSystem, Payload System and Energy System. UUV control system involvesautonomous controlsystem and motion control system, including integrated application ofintelligent architecture,autonomous navigation, task planning, motion control and other subsystems.Autonomous controlsystem is the core and key system for UUV to autonomously complete operationaltasks. It isequivalent to human brain playing the function of intelligent planning anddecision-making basedon sensory information, especially for UUV, is a key composition as a kind ofunmanned systemswithout human guidance. Furthermore, this invention proposes the targettracking system based onLong Short-Term memory (LSTM) neural network. The target tracking system isimplemented inautonomous control system. Finally, the target tracking system is verified onan unmanned mobilerobot. The target tracking system based on LSTM neural network belongs to thefield ofunmanned underwater navigation. The invention uses LSTM neural network totrack the target ofcomplex and non-linear target. It solves the problems of difficult tracking ofmaneuvering targets,difficulty in establishing target models, and low tracking accuracy. First, wewill collect thelatitude and longitude information and velocity information of the target andperform dataprocessing. Then, we design the LSTM neural network structure for singletarget tracking. Finally,LSTM neural network parameters will be adjusted optimally to achieve targettracking. Theinvention effectively simplifies the non-linear filtering process and caneffectively track complexnon-linear targets. It does not require the establishment of a target modeland the use of traditionalfiltering algorithms. The target's historical information will be used toestimate the target'smovement status at the next moment. At the same time, the neural network usesthe internalparameters adjusted by the back-propagation algorithm and the attenuation ofthe learning rate toreduce the amount of calculation. This target tracking system is designed forunmanned vehiclesand UUVs.2122