The invention discloses an unmanned vehicle local path
planning method based on a neural network and a dynamic window method. The method comprises the steps of establishing a
data set and dividing training and testing sets; performing iterative optimization on the
residual neural network by using the
training set; evaluating by using the
test set to obtain an optimized
residual neural network; a map information matrix is obtained and input into the optimized
residual neural network for path planning, and a corresponding
azimuth angle planning action is obtained; the corresponding optimal
angular velocity and the optimal linear velocity are obtained, and then the unmanned vehicle moves to the next position; and completing navigation and the like. According to the method, the next action of the unmanned vehicle is planned through the neural network, the problem that
local optimum and narrow space are difficult to process is solved, and target
accessibility in a complex environment is guaranteed; the unmanned vehicle is actually controlled to move through a dynamic window method, and absolute safety of
obstacle avoidance of the unmanned vehicle is guaranteed; the
system has an autonomous learning ability and can adapt to various different working environments; and direct planning in the velocity space is also convenient for direct actual deployment.