Research On Adaptive Local Risk Avoidance Of USV

Risk avoidance technology is one of the most important researches on unmanned surfacevessel currently. During the process of missions which the USV commits on, the USV need tobe highly intelligent and avoid the obstacles in the sea adaptively. There are two kinds of riskavoidance technology: the global risk avoidance and the local risk avoidance. To achieve theglobal risk avoidance, we usually take the method of static planning; During the sail of theUSV in high speed, it is much more important for the USV to take the process of local riskavoidance. So, it is meaningful to take the reach on the adaptive local risk avoidance of theUSV.This article mainly takes the study on the ways of local risk avoidance of the USV whichis sailing in the complex marine and during the process of local risk avoidance, the USV needto adapt to the impact of the sea breeze. The marine is complex and changeable, the factors ofthe marine which affect on the USV are the wind, waves and the currents and so on. Takinginto account of the character of high speed, during the navigation process of risk avoidance ofthe USV, the impact is particularly significant which is brought by the sea breeze. To ensureits own safe, the USV needs to realize risk avoidance reliably. This article introduces themechanism of the reinforcement learning to further improve the self-adaptive of the USV. Themechanism of the reinforcement learning is realized by the Q-learning which is combinedwith the chaos theory. The main tasks which are completed in the article are:This article introduces the development status and the trends of the USV domestic andabroad first, and then points out the hot relevant research focus, and proposes the researchbackground and significance lastly. After that, this article analyzes the influence impacted onthe local risk avoidance of the USV which is brought by the complex marine, and introducesthe research status and results of the local risk avoidance methods home and abroad. What isnext is that we propose an improved local risk avoidance method of the USV which is calledfuzzy ND method. The point of this article is researching the method of local risk avoidanceof the USV take consider of the sea breeze impact. In order to increase the adaptive capacityof the local risk avoidance of the USV, this article adopts the idea of learning, and introducesthe Q-learning mechanism. The design process of the Q-learning is represented in the article, we use Q table to realize the Q function and adopt chaos theory combines withε-greedy todeal with the action selection and using problem during the process of learning. In order toimprove the efficiency and the convergence rate of the Q-learning method, this articleproposes the concept of similar wind state based on the characters of the sea breeze，updatesmulti-Q-values at the same time using the internal relations of the similar wind state.Simulation results verify the feasibility of the method mentioned above, and the simulationresults are analyzedFinally, a summary of the full text is made, and future work which need to be done isdiscussed.