A sea-air cross-domain multi-agent hierarchical cooperative task planning system

By employing a hierarchical collaborative mechanism and a multi-agent reinforcement learning algorithm, the problems of high task conflict rate, poor environmental dynamic adaptability, and high computational complexity in cross-domain sea-air collaborative planning are solved, achieving efficient and rapid cross-domain multi-agent task planning.

CN122195097APending Publication Date: 2026-06-12CHENGDU SIDU SPACE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHENGDU SIDU SPACE TECH CO LTD
Filing Date
2026-05-18
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

Existing technologies suffer from high task conflict rates, poor environmental adaptability, and high computational complexity in cross-domain collaborative planning between sea and air, especially in large-scale intelligent agent collaborative scenarios, making it difficult to achieve efficient and rapid task planning.

⚗Method used

A hierarchical collaborative mechanism is adopted, including a global task planning layer, an intra-domain coordination layer, an agent execution layer, and an environmental perception layer. It combines improved genetic algorithms, ant colony algorithms, A* algorithms, and artificial potential field methods with the QMIX multi-agent reinforcement learning algorithm to achieve collaborative task planning of multi-agent agents across sea and air domains.

🎯Benefits of technology

It effectively reduced the task conflict rate to ≤5%, shortened the planning adjustment time to ≤2s, improved task completion efficiency by 22%, and supported collaborative planning of 50 agents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122195097A_ABST
    Figure CN122195097A_ABST
Patent Text Reader

Abstract

The application discloses a kind of sea-air cross-domain multi-agent layered collaborative task planning system, belong to agent cooperation and task planning technical field.The four-layer architecture design is used: global task planning layer is realized task decomposition and resource allocation using improved genetic algorithm;Domain coordination layer plans unmanned aerial vehicle and unmanned ship path respectively through ant colony algorithm (air) and A* algorithm (sea surface);Artificial potential field method is realized local obstacle avoidance in agent execution layer;Environment perception layer integrates high-precision sensor to monitor environment in real time.The system uses QMIX multi-agent reinforcement learning algorithm, and realizes agent cooperation through positive and negative reward mechanism (task progress reward+10 / 10% progress, environment inadaptation reward-20 / wind speed≥15m / s).The application supports the cooperation of more than 50 agents, the task conflict rate is less than or equal to 5%, the planning adjustment time is less than or equal to 2s when the environment changes, the task completion time is shortened by more than 20%, and the cooperation efficiency and reliability of sea-air cross-domain task are significantly improved.
Need to check novelty before this filing date? Find Prior Art