Research On Path Planning Based On Markov Decision Process For AUV

The autonomous path planning is an important ability for Autonomous underwater vehicle (AUV), which is acrucial factor for ensuring and guiding AUV to accomplish its long-termnavigating and autonomous mission in large range and wicked oceanic area. Some technologies are all deeply researched in the dissertation, that are the global path planning technology in known and certain environment, the local path planning technology in unknown and uncertain environment.According to the theory of Markov Decision Process, an algorithm is developed and applied to solve the problem of AUV 2-D global path planning in complex ocean environment. The model of path planning is designed, and the program of path planning is regarded as looking for the best tactical through the model of entironment and principle of rewards and punishment. Introducing the theory of MAXQ, the papar designed the hierarchical assignment system for AUV global path planning, and made the planning more reasonable. Furthermore, the current in ocean is discussed, and correspondingly the simulation test is done for AUV with even current or varing current. The ability of anti-current for AUV is built up.For the problem of dimension disaster, the paper present a hierarchical MDP based on the method of clustering state. A matching from probabilities assignment of hierarchical states to the optimal action for AUV is realized. Through the comparing of two simulation data, we can see the hierarchical MDP have some merits, such as less nodes and cost, and it can reduce the times of changing heading for AUV.In addition, a novel idea of spatial model based on hierarchical depth partition is present to avoid the confliction between computer storage capacity and large-scale data.For the unknown environment, based on the vision model of FLS, a local planning method is discussed. Particularly, the paper designed the Partially Observable Markov Decision Process model for local path planning. The collision risk model of underwater obstacle is set up for the first time. The judgment criterion of obstacle motion property and the linear motion predictor of dynamic obstacle are proposed. As an emphasis, sloving the problem of danger degree model for velocity protential, and joined velocity factor to the POMDP model. Simulating test is given for velocity planning. Using the semi-physical simulation system for path planning, the arithmetic above all is validatd in simulation test.To simulate the 3-D motion and avoiding obstacles of AUV in ocean environment, the simulation platform of space avoiding obstacles for AUV is built. Using the laser measure system to simulate the sonar of AUV, it is good at simulating the avoiding obstacles for AUV in complex environment, and detecting the unkown environment. It made the foundation for real AUV of avoiding obstacles in ocean.By the autonomous planning technology, the adaptive and autonomous abilities of AUV can be enhanced and improved in large-range and complex oceanic environment. And the safety, reliability and efficiency requirement of AUV can be met with the need of long-term navigating and seabed survey mission. It has important and considerable value whether in theory research or in practice application.