Research On Sensing And Decision-making For Navigation Threat Of UUV

The autonomous ability of UUV which is the key competitive ability has become a topic research branch in ocean exploitation. The security of UUV is one of the most significant branches of autonomous ability. Like a person, security of UUV is the precondition of all the action. The security threat encountered of UUV include self and environment threat. Self-threat mainly means the failure, and obstacle is the most troubled threat in the unknown environment. The paper, considered the security, is established in the improvement of autonomous ability. SVM and fuzzy theory is applied to research the thruster failure detection and the obstacle avoidance.Aimed at self-threat, the paper works over the thruster failure detection which is the most minatory threat. The armature current of thruster has memory and this peculiar property induces the hysteresis. As a result, existing current is influenced by prior current. The paper proposes the forecasting current, which is calculated by the historical current using SVM forecasting, and the result is credible. The theory current gained by fitting relation is proposed to incarnate the influence of control voltage. Fusion coefficient is dynamically adjusted according to absolute value of control voltage derivative and the anomaly current time interval. The paper designs the fuzzy rules and membership function. Based on the fusion current, thruster failures of UUV can be detected and classified in real time. Through the lake experiment and sea trials, the algorithm in this paper has neither brought "false alarm", nor has accomplished the accurate diagnosis of propeller fault. The fault classification provides the basis for fault-tolerant control.To reduce the environment threat of UUV, an avoiding system including obstacles data processing and the obstacle avoidance behavior is proposed. The UUV in a trial is equipped with three forward-looking range sonar. The virtual sonar wave is proposed to make up the lack of information for range sonar. The virtual sonar wave is computed by the obstacle map. Comparing to radar and ultrasonic used in air, the range sonar is not credible and easily disturbed. The clustering application based on SVM is proposed to process obstacle data. Through clustering, outlier can be detected and eliminated. Thus the obstacle with arbitrary shape can be detected, which is the basic for avoiding decision. At last, obstacle avoiding behavior based on fuzzy theory is designed. The entire obstacle avoiding system is a typical fuzzy system, which divides into4fuzzy subsystems. According to motion characteristic and the unknown condition, the fuzzy rules and membership function is designed. Simulation and lake trials show the validity of the method presented that it can eliminate the outlier points and avoid obstacles successfully. The vehicle arrives at the aim point.