Deep Learning-based Decision-Making Model for the Submarine Evade Movement

It is of great significance to study the intelligent countermeasure decision of submarine eluding torpedo for the successful defense precision of underwater acoustic warfare. In this paper, the problem in view of the traditional submarine evasion decision-making which excessively relies on the previous combat experience, a deep learning-based decision-making model is proposed for the submarine's evasive movement. Submarine evasion confrontation model and underwater acoustic antagonism system (UAAS) are used to establish a depth neural network (DNN) which with initial course, speed and depth of torpedo and submarine as input predict the best evaded course, speed and depth of submarine. Taking four different battlefield situations as examples, the simulation analysis shows that the model can increase the success rate of submarine evasion by about 19.7% compared with the randomly selected evasion decision which conforms to the actual attack, which indicates that the model has high accuracy and strong feasibility, and it provides a reference for the intelligent decision of submarine underwater acoustic countermeasure.