Mobile robot visual navigation method and device based on deep reinforcement learning

Abstract

The invention relates to the field of machine vision navigation, in particular to a mobile robot vision navigation method and device based on deep reinforcement learning, and the method is based on a deep reinforcement learning method, takes an image, a depth image and a target position as input, and can realize navigation in a large space with multiple mixed scenes. The navigation capability of a mobile robot visual navigation technology based on deep reinforcement learning is improved; in addition, by designing the speed of the mobile robot and a reward function related to the distance between the mobile robot and the target, the training of the deep reinforcement learning model can be quickly converged; according to the method, the navigation capability of the deep reinforcement learning method in a complex large scene can be improved, the problem of sparse rewards is solved, the model convergence speed is accelerated, and the navigation performance in the complex large scene is improved.

Description

technical field [0001] The present invention relates to the field of machine vision navigation, in particular to a method and device for visual navigation of a mobile robot based on deep reinforcement learning. Background technique [0002] The visual navigation method of mobile robots in complex and large scenes based on deep reinforcement learning takes the currently observed image and target information as input, and outputs continuous actions to make the agent avoid obstacles and take a short path to the designated position. At present, the visual navigation technology of mobile robots based on deep reinforcement learning mainly has the following two problems: one is that the visual navigation performance in a large space is weak, and the other is that it is difficult to navigate in a variety of different scenarios at the same time. [0003] At present, the positioning technology that is relatively mature and widely used is the global positioning system positioning techn...

AI Technical Summary

Problems solved by technology

At present, the visual navigation technology of mobile robots based on deep reinforcement learning mainly has the following two problems: one is that the performance of visual navigation in large spaces is weak, and the other is that it is difficult to navigate in multiple different scenes at the same time.

However, this kind of technology will have problems such as weak signal and inaccurate positioning indoors, resulting in poor navigation effect, and the goal of simultaneously navigating indoors and outdoors cannot be achieved; Technology and technology based on visual navigation, complete simultaneous positioning and mapping through laser radar and visual sensors, and realize the navigation of mobile robots; have a clear understanding of the environment

[0004] The existing technology does not fully integrate the depth information in the image, the generalization ability and obstacle avoidance ability of unknown scene objects are poor, and the reward function design for deep reinforcement learning is relatively simple, and the problem of reward sparsity is very easy to occur, resulting in mobile robots. It is extremely difficult to reach the target point, resulting in slower training convergence speed, and the navigation performance is greatly weakened in complex large spaces

Images