Research On Self-repair Technology For AUV Software Based On POMDP

With the growing strength of the human exploration of the oceans to increased,as the main tool of marine research and development-AUV(Autonomous Underwater Vehicle)is widely used in deep-sea waters and a variety of high-risk jobs,in this highly complex marine environment AUV equipped mainly relying on its own intelligent control type of software to perform the job,once the AUV its software faults and how to ensure that it can be repaired timely has become an important research topic.Aiming at the AUV intelligent decision software problem caused by the system of aging and system scheduling anomaly,this article build a set based on POMDP(Partially Observable Markov Decision Process)three layer model of micro reboot and self repair system,ensure that intelligence,timely and efficient repair of AUV software of abrupt faults.This article realize a solution for AUV software failure problems,the solution using in the three micro-reboot as troubleshooting tools,build self-repair POMDP model used to generate the optimal repair strategy.Firstly,through failure mode effects analysis of AUV intelligent decision-making software and software platform Vx Works,design a modular architecture of AUV software fault repair;ensure that this structure can repair the software failure automatically.Secondly,in order to solve the traditional troubleshooting method of high cost and low efficiency,this subject adopts micro-reboot technology to build troubleshooting methods with low cost,and combining with the characteristics of AUV software hierarchical write design based on three layers of micro restart the realization form,this form make repair action selection more flexible and repair effect more efficient.Then,in AUV software troubleshooting strategies,this topic for the AUV software failure state part perception build POMDP self-repair model,using learning method of based on historical data to generate model parameters,and the PBVI algorithm is introduced into the self-healing strategy,while improve PBVI algorithm applied to online strategy for solving the self-healing,to ensure that self-repair system can according to the fault problem because of AUV software aging and abnormal operation automatically determine optimal repair scheme.Finally,considering self-repair model parameters based on historical data are not well fit real failure environmental issues when environment changes,design online self-repair model parameterslearning algorithm to ensure that self-repair system can adapt to changes of the real environment.In order to verify the overall feasibility of self-repair program,this article through simulation experiment verify the converges and real-time of the self-healing strategy algorithm,and verify the accuracy of fault repair automated processes,verify the optimality and stability of fault repair effects,while analysis on the main factors of affecting self-repair effect based on the experimental data.Experimental results show that the self-repair program for AUV software failures caused by aging and system calls timely select optimal repair and restoration action better,for the sudden failure occurring during the processing of AUV underwater operation have good research and popularization value.