A mobile robot autonomous exploration method based on deep reinforcement learning

By adopting deep reinforcement learning methods in the autonomous exploration of mobile robots, building a reinforcement learning strategy network and designing optimized reward functions, the problem of repeated exploration and high computing resources consumed by mobile robots in unknown environments is solved, and more efficient and stable autonomous exploration is achieved.

CN119200601BActive Publication Date: 2025-03-28GUANGZHOU NENGXIN DIGITAL TECH CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202411312042.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2025-03-28
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

In the prior art, the autonomous exploration method of mobile robots is prone to repeatedly explore known areas, resulting in low efficiency, and large-scale map construction consumes a lot of computing resources and memory, and relying on environmental models leads to poor stability of search strategies in real scenarios.

Method used

Using the autonomous exploration method of mobile robots based on deep reinforcement learning, by building a reinforcement learning strategy network, designing an optimized reward function, considering the distance and angle differences between the robot pose and the next waypoint, and based on the curiosity of the explored and unexplored areas, reduce repeated exploration and improve exploration efficiency.

Benefits of technology

It greatly reduces the situation where mobile robots repeatedly explore known areas, improves the efficiency and stability of independent exploration in unknown environments, and reduces the consumption of computing resources and memory.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a method for autonomous exploration of a mobile robot based on deep reinforcement learning, the method comprising: turning on a laser radar device, a computing unit and a robot body power supply in an unknown environment, collecting point cloud data of the current environment through the laser radar; storing the point cloud data based on a complementary hole structure; constructing an information graph based on a two-dimensional occupancy map, and sparsely processing the information graph; constructing a reinforcement learning strategy network, and designing an objective function of the reinforcement learning strategy network; a local planner generates a speed value of the current robot according to a collision-free path; a motion controller of the robot generates a motion command according to the speed value, and repeating the above steps until the autonomous exploration task of the robot based on deep reinforcement learning is completed; the advantages of the present invention are: using model-free deep reinforcement learning technology to learn exploration strategies in a simulation system, which can be seamlessly transplanted to a real physical environment, and can achieve more efficient autonomous exploration.
Need to check novelty before this filing date? Find Prior Art