Research Of Multi-AUV Task Allocation Based On Robust Restless Bandit Formulation

The ocean had been concerned more and more by many countries around the world,the number of the research and industry of ocean is increasing in recent years.The research and industry about ocean are all based on the underwater information in it,there is a key problem that how to acquire the underwater information.Because of the special of underwater environment,radio signal is attenuated severely,so the wireless sensor network technology can't use directly,and the underwater acoustic monitoring network would be a good choice to acquire underwater information.In this passage,we consider a new acoustic monitoring network,which,instead of the original gateway node,Autonomous Underwater Vehicles(AUV)are used to acquire underwater information,so the complexity of the network is reduced,and the reliability and robustness is improved.The main problem what we researched in this passage is get the effective task allocation policy of AUV to acquire underwater information,achieve the underwater acoustic monitoring network,and against the complex hard underwater environment.Firstly,we introduced the underwater acoustic monitoring network system what we considered,and analyzed the influencing system factors such as channel noise character in ocean.We also introduced the sequential decision model and the Restless Bandit Problem model,include model parameters and structure,the solution of restless bandit problem is also analyzed.Secondly,the model of AUV task assignment system based on restless bandit in this paper is introduced.Consider the time-varying property of underwater channel quality parameter and its influence,we also modeled the dynamic of channel quality as a process of restless bandit.Then we mixed the channel quality model and sensor model,achieve a mixed RBP model of AUV task assignment system,so we can consider the both influence of 2 factors simultaneously.For the problem of solve RBP model,we raise a Whittle Index method which suited our problem,and get the AUV allocation policy.Finally,we considered the problem of the imprecision of AUV task assignment RBP model because of the multi-interference factors underwater.An uncertain parameter is added to the RBP model,so we can calculate the robust RBP model of AUV task assignment system,and solve the robust AUV allocation policy out.By getting the uncertain parameters for each sensors and channel respectively,and setting the penalty values for them,we can calculate the robust RBP model accurately.Then we can solve out the robust AUV allocation policy by the robust model based on the rectangular uncertainty characteristics.The accuracy of system model is improved and the efficiency of AUV allocation policy is increased by this method.