Research on Autonomous Collision Avoidance Method of Unmanned Surface Vessel in the circumstance of Moving Obstacles

This paper researches autonomous collision avoidance method of USV in the circumstances of moving obstacles, using the approach that based on the prediction model of Elman network to estimate motion state of the moving obstacle. The state of moving obstacle can be predicted relaying on the available movement position data measured. The internal feedback signal is added to describe the nonlinear dynamic problem, and obtaining the information of the obstacle closest to the actual state of motion; Using the improved artificial potential field algorithm, it is applied to the USA collision avoidance system, taking into account the relative distance between the USA and the destination point as well as the estimated collision time. As a result, the solution of field function does not fall into the local minimum, avoiding the situation which the USV is in local oscillation or loss the path to destination when close to the destination point. Finally, the correctness of the proposed method is verified by simulation.