多智能体集群协作搜救系统 (Multi-Agent SAR System)

Multi-Agent Cooperative Search and Rescue System

这是一个面向复杂非结构化环境的无人机集群协作搜救仿真系统。 This is a UAV swarm cooperative search and rescue simulation system designed for complex unstructured environments.

本项目深度集成了机器人控制理论、信息论以及多智能体系统 (MAS) 领域的多种前沿算法，构建了一个具备自主感知、动态决策、路径规划和分布式协作能力的智能体集群仿真平台。 This project deeply integrates cutting-edge algorithms from robot control theory, information theory, and Multi-Agent Systems (MAS) to build an agent swarm simulation platform with autonomous perception, dynamic decision-making, path planning, and distributed collaboration capabilities.

核心特性

Core Features

• 先进避障算法: 采用 涡旋势场 (Vortex Potential Field) 算法，有效解决了传统人工势场法 (APF) 的局部极小值问题，实现密集障碍物环境下的平滑绕行。

  Advanced Obstacle Avoidance: Adopts the Vortex Potential Field algorithm, effectively solving the local minima problem of the traditional Artificial Potential Field (APF) method and achieving smooth navigation in dense obstacle environments.

• 信息驱动覆盖: 实现基于 分布式 Voronoi 覆盖控制 与 信息熵地图 的结合，引导集群优先探索高不确定性区域。

  Information-Driven Coverage: Implements a combination of Distributed Voronoi Coverage Control and Information Entropy Maps to guide the swarm to prioritize exploration of high-uncertainty areas.

• 高效搜索策略: 引入 Lévy Flight (莱维飞行) 机制，利用重尾分布的随机游走策略，平衡局部搜索与广域探索，显著提升稀疏目标环境下的搜索效率。

  Efficient Search Strategy: Introduces the Lévy Flight mechanism, utilizing a heavy-tailed random walk strategy to balance local search and wide-area exploration, significantly improving search efficiency in sparse target environments.

• 异构角色协作: 支持 侦察者 (Scout)、确认者 (Verifier) 和 支援者 (Support) 等角色的动态分配与调度。

  Heterogeneous Role Collaboration: Supports dynamic assignment and scheduling of roles such as Scout, Verifier, and Support.

• 高保真感知模型: 摒弃简单的二元感知，构建了包含视场角 (FOV)、距离衰减和视线遮挡 (Occlusion) 的 概率感知模型。

  High-Fidelity Perception Model: Discards simple binary perception and builds a probabilistic perception model incorporating Field of View (FOV), distance attenuation, and Line of Sight (Occlusion).

系统架构

System Architecture

系统遵循 OODA (Observe-Orient-Decide-Act) 循环架构： The system follows the OODA (Observe-Orient-Decide-Act) loop architecture:

1. 环境层: 维护地图数据、障碍物分布及全局信息素。 Environment Layer: Maintains map data, obstacle distribution, and global pheromones.
2. 感知层: 模拟传感器特性，处理视线遮挡与概率检测。 Perception Layer: Simulates sensor characteristics, handling occlusion and probabilistic detection.
3. 决策层: Decision Layer:
   • 指挥中心: 基于拍卖算法 (Auction Algorithm) 进行全局任务分配。 Command Center: Performs global task allocation based on the Auction Algorithm.
   • 智能体: 基于有限状态机 (FSM) 进行局部决策与控制。 Agents: Perform local decision-making and control based on Finite State Machines (FSM).
4. 执行层: 处理动力学约束与电量管理。 Execution Layer: Handles dynamic constraints and battery management.

项目结构

Project Structure

.
├── src/                          # 源代码目录 (Source Code Directory)
│   ├── vortex_field.py           # 涡旋势场算法 (Vortex Potential Field Algorithm)
│   ├── drone_agent.py            # 无人机智能体逻辑 (Drone Agent Logic)
│   ├── coordinator_agent.py      # 指挥中心逻辑 (Coordinator Logic)
│   ├── environment.py            # 环境与物理规则 (Environment & Physics)
│   └── ...
├── figures/                      # 结果图表目录 (Results & Figures)
├── simulation.py                 # 仿真主程序 (Simulation Main Entry - GUI/Text)
├── visualization.py              # 可视化渲染模块 (Visualization Module)
├── run_experiments.py            # 批量实验脚本 (Batch Experiment Script)
├── plot_results.py               # 数据绘图脚本 (Data Plotting Script)
├── requirements.txt              # 项目依赖 (Project Dependencies)
└── README.md                     # 说明文档 (Readme)

快速开始

Quick Start

1. 环境安装

1. Installation

本项目基于 Python 3.8+ 开发。 This project is developed based on Python 3.8+.

pip install -r requirements.txt

2. 运行仿真

2. Running Simulation

启动 GUI 可视化模式: Launch GUI Visualization Mode:

# 基础场景 (Basic Scenario)
python simulation.py --scenario medium

# 密集障碍物场景 (观察涡旋避障)
# Dense Obstacle Scenario (Observe Vortex Avoidance)
python simulation.py --scenario hard --drones 4

# 城市覆盖场景 (观察 Voronoi 覆盖)
# City Coverage Scenario (Observe Voronoi Coverage)
python simulation.py --scenario city --drones 6

启动无界面文本模式: Launch Text-Only Mode:

python simulation.py --text --scenario medium

3. 实验数据生成

3. Generating Experiment Data

如需复现性能分析图表（避障对比、覆盖率分析等）： To reproduce performance analysis charts (obstacle avoidance comparison, coverage analysis, etc.):

# 运行批量实验 (耗时较长)
# Run batch experiments (Time consuming)
python run_experiments.py

# 生成统计图表
# Generate statistical charts
python plot_results.py
python generate_trajectory_plots.py