Unmanned vehicle navigation method based on deep reinforcement learning

Abstract

The invention provides an unmanned vehicle navigation method based on deep reinforcement learning. The method comprises steps of obtaining a depth image through a depth camera on an unmanned vehicle, carrying out the sampling of the obtained depth image, and carrying out the quadratic linear interpolation processing to form a depth image matrix; in the depth image matrix, relative positioning with the starting point being calculated through a wheel speed odometer of the unmanned vehicle, and a second depth image matrix representing the state of the unmanned vehicle being formed; comparing numerical values in the second depth image matrix one by one, calculating a minimum value of certain values, comparing the minimum value with a set threshold value, controlling the movement of the unmanned vehicle in a kinematics mode when the minimum value is greater than the set threshold value, and inputting the second depth image into a deep learning network when the minimum value is smaller than the set threshold value, the next action being determined randomly or according to a deep learning network. According to the method, the network learning efficiency is higher, the error convergence value is smaller, the obstacle avoidance effect of an unknown environment is better, and map acquisition efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of unmanned vehicles, in particular to a navigation method for unmanned vehicles based on deep reinforcement learning. Background technique [0002] The navigation of unmanned vehicles is a variety of technologies to avoid obstacles and reach the target position. However, navigating in an unknown environment is much more difficult than a known environment. When the environment is unknown, the movement of the robot is extremely dependent on the information collected from the sensor. The efficiency of data and algorithms to find a good path, the sensors on the mobile robot will help detect obstacles, and draw a map of the environment when navigating to navigate to the target location. When the deep reinforcement learning method is applied to the research of navigation problems, a successful training model is highly dependent on the training set information, so it is inevitable that a lot of time is needed to c...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of the present invention provides a navigation method for unmanned vehicles based on deep reinforcement learning to solve the problem that the existing unmanned vehicle navigation methods have poor adaptability to the environment and poor universality from the training environment to another unknown environment , a technical problem requiring a long training time

Images