The invention discloses an
underwater vehicle cooperative detection method based on multi-agent
reinforcement learning, and the method comprises the steps: obtaining the original observation flow,
ocean current field data and sound velocity profile data of an
underwater vehicle, and generating a
time sequence enhancement state containing the
kinematics estimation of the
underwater vehicle and underwater acoustic link
delay distribution; calling a pre-trained strategy model based on the
time sequence enhancement state, and outputting a constraint multiplier vector; according to the constraint multiplier vector, in combination with an energy upper limit constraint, a
signal-to-
noise ratio threshold constraint and a minimum safety distance constraint, executing lower-layer optimization, and generating a feasible planning solution containing a path set and
connectivity configuration; based on the feasible planning solution, calling a pre-trained commentator model, and accounting communication value evaluation; and deciding whether to execute
underwater acoustic communication or not according to the communication value evaluation and the communication triggering threshold. According to the method, the technical problems of representation failure, difficulty in
reinforcement learning and complex physical constraint decoupling and low communication
resource efficiency in a high-
delay asynchronous state are solved.