Robot post-disaster rescue collaboration method based on deep learning and large language model

CN122196858APending Publication Date: 2026-06-12李祥健
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
李祥健
Filing Date
2024-12-12
Publication Date
2026-06-12

AI Technical Summary

Technical Problem

Existing disaster relief robot systems have significant shortcomings in terms of rigid task scheduling, insufficient multimodal perception, and unstable collaborative communication. They are unable to cope with the high dynamism and complexity of the post-disaster environment, resulting in resource waste, task delays, and operational failures.

Method used

By employing a deep learning and large language model-based approach, through task semantic graph generation, multimodal data fusion, distributed perceptual bridge, and reinforcement learning model, we can achieve dynamic adjustment of task priorities, information sharing, and optimization of robot collaboration strategies. We can also optimize robotic arm operations by combining force perception and semantic understanding.

Benefits of technology

It improves the flexibility and adaptability of robots in disaster relief scenarios, enhances the accuracy and timeliness of information sharing, improves the efficiency and success rate of task planning, and ensures collaborative capabilities in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122196858A_ABST
    Figure CN122196858A_ABST
Patent Text Reader

Abstract

The application discloses a kind of robot post-disaster rescue collaboration methods based on deep learning and large language model, by introducing " reflection mechanism " in natural language processing, realize task semantic analysis, dynamic optimization and adaptive scheduling.Large language model generates task semantic graph and combines reinforcement learning to adjust priority in real time, form closed-loop task reflection and optimization process, greatly improve the flexibility and adaptability of robot in complex environment.Design " distributed perception bridge " module, through attention mechanism dynamic screening and sharing key perception data, realize the efficient cooperation of heterogeneous robot, even in the condition of unstable communication still can guarantee rescue efficiency.Based on the reinforcement learning model driven by environmental characteristics, real-time perception of physical properties such as fragment density and slope, optimize behavior strategy, improve the generalization ability and robustness of model.Through " force-semantic coupling " motion planning, dynamically adjust the operation path and force of manipulator, improve the success rate of complex task.
Need to check novelty before this filing date? Find Prior Art