Autonomous Behavior-learning And Planning Of AUV Space Motion

Autonomous Underwater Vehicle (short for AUV) is a new concept of unmannedplatforms. Through carrying the different sensors and the mission modules, It cancomplete the underwater observatory, marine development, marine engineering and othertasks. In this paper, for spatial characteristics of the motion of AUV, an autonomousagent based on behavior and dynamics method is proposed. This article deeply studies themotion planning of AUV Autonomous Behavior Agent.In this paper, several aspects are studied as follows:In this paper, for the motion planning problem in the complexity of the marineenvironment, the AUV autonomous agent is established based on behavior and dynamicsmethod. It establishes independent horizontal independent behavioral model of thedynamics of behavior-based AUV. According to fuzzy theory, the behavioral models ofvertical fixed-deep navigation behavior and fixed-high navigation behavior areestablished. Combining behavior fusion mechanism whit Petri nets of the discrete eventdynamic systems, the behavior-task coordination mechanisms have been proposed.Simulation results show that: the overall behavior of coordination mechanisms of theAUV is reliable and reasonable, and be able to successfully coordinate behavior on theAUV horizontal and vertical movement.The algorithm of this article has been improved to the poor generalization ability ofthe shortcoming of traditional reinforcement learning algorithm. In this paper, using theneural network Q-learning algorithm, the generalization of the AUV learning system hasbeen greatly improved. Applied to the independent behavior of the AUV learningplanning, self-learning agent is established based on neural network Q-learning for AUV.The simulation results show that the AUV can achieve its motion planning through thelearning experience of independent study in unknown environment. Adequately verifycorrectness of theAUV self-learning and its generalization ability.In order to validate that the built AUV autonomous behavior-learning agent isreliable and reasonable, the simulation experiment platform of the AUV intelligent agentis established. And design test cases of the independent capacity and self-learning abilityof the AUV intelligent agent. The results of test cases show that the built AUVautonomous behavior-learning agent can successfully complete a variety of sea-crossing and job tasks.