Multi-target hunting method and system for mobile robot based on density field and deep reinforcement learning

By combining density field and deep reinforcement learning, the problems of insufficient dynamic adjustment capability and high training complexity in multi-target capture of mobile robots are solved, achieving efficient, uniform and fast multi-target capture, and improving system performance and scalability.

CN120124971BActive Publication Date: 2025-11-28SOUTHEAST UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202510274134.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-03-10
Publication Date
2025-11-28
Estimated Expiration
2045-03-10

AI Technical Summary

Technical Problem

Existing multi-target capture methods for mobile robots suffer from low capture efficiency due to insufficient dynamic adjustment capabilities, while learning-based methods face the problems of high training complexity and difficulty in convergence when multiple targets are captured.

Method used

This method combines density field analysis and deep reinforcement learning. Through seven steps—perceptual information acquisition, target density calculation, target allocation decision, encirclement state judgment, observation state generation, and action decision and execution—the density field algorithm is used to calculate the target density value and combine it with the relative position to calculate the score. This enables reasonable real-time allocation of targets and robot grouping. The robot then performs efficient encirclement based on the strategy trained by deep reinforcement learning.

Benefits of technology

It significantly improves the encirclement effect, achieves uniform and rapid multi-target encirclement, reduces training complexity, improves the system's versatility and real-time performance, and adapts to multi-target encirclement tasks of different scales.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120124971B_ABST
    Figure CN120124971B_ABST
Patent Text Reader

Abstract

The application discloses a kind of mobile robot multi-target hunting methods and systems based on density field and depth reinforcement learning, adopt density field to combine depth reinforcement learning technology, including seven steps of perception information acquisition, target density calculation, target allocation decision, hunting state judgment, observation state generation, action decision and execution and hunting process iteration, score is calculated using density field algorithm in combination with relative position, to realize the reasonable real-time allocation of target and robot grouping, ensure that hunting force is balanced;The strategy of robot in the same group based on depth reinforcement learning training is efficiently hunted to target.The method of the application significantly improves the system performance in the aspects of mobile robot quickly approaching target and quickly forming hunting formation, and provides an efficient and intelligent solution for multi-target hunting scene.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Multi-robot cooperative tracking method for overwater hunting task

    CN117289698A

  • Neighbor distribution condition-based cluster robot surrounding target selection method

    CN118884950A