AI for Controlling a Simulated Swarm of UAVs with Pollutant-Cloud Tracking

This project implements an AI system to control a swarm of unmanned aerial vehicles (UAVs) for tracking and monitoring pollutant clouds. The simulation includes realistic UAV dynamics, UAV-to-UAV communication, battery life limitations, and cloud dispersion models.

Project Overview

The system simulates a swarm of UAVs that work together to track and monitor the concentration of pollutants in a moving cloud. Each UAV is equipped with:

• Sensors to detect pollutant concentration
• GPS for navigation
• Communication capabilities to coordinate with other UAVs
• A limited battery life

The UAVs use a combination of swarm intelligence algorithms to efficiently track the cloud while avoiding collisions with each other.

Features

• Realistic UAV dynamics with velocity and turn rate constraints
• Simulated sensor noise and GPS error
• Inter-UAV communications with message delay
• Collision avoidance algorithms
• Pheromone-based coordination strategy
• Battery life simulation with random failures
• Two different cloud dispersion models for testing
• Easy-to-use configuration interface

Files and Components

Core Simulation Files

• sim_start.m: Main simulation script that runs the simulation
• UavBrain.m: Implements the AI decision-making logic for each UAV
• UavBody.m: Physical model of the UAV with movement constraints
• UavState.m: Defines the states for the UAV state machine
• Pheremone.m and PheremoneType.m: Implements the pheromone-based coordination system
• Message.m: Defines the communication protocol between UAVs

Cloud Simulation

• cloudplot.m: Visualization for the pollution cloud
• cloudsamp.m: Sampling function for pollution concentration
• cloud1.mat and cloud2.mat: Pre-computed cloud dispersion data

Movement Functions

• moveRungeKutta.m: Advanced movement calculation
• forwardEuler.m: Simple movement calculation
• sqrLen.m: Utility function for distance calculations

User Interface Scripts

• run_simulation.m: Easy-to-use script for configuring and running simulations
• visualize_results.m: Post-simulation analysis and visualization tool

Requirements

• MATLAB R2014b or later
• No additional toolboxes required

How to Run

There are two ways to run the simulation:

Method 1: Using the Configuration Script (Recommended)

1. Open MATLAB
2. Navigate to the project directory
3. Open and edit the parameters in run_simulation.m as desired
4. Run the script:

   run_simulation

Method 2: Running the Simulation Directly

1. Open MATLAB
2. Navigate to the project directory
3. Run the main simulation script with default parameters:

   sim_start

Configuration Options

The following parameters can be adjusted in run_simulation.m:

• uavCount: Number of UAVs in the swarm (default: 24)
• dt: Time step for simulation (default: 1.0 second)
• maxTime: Maximum simulation time (default: 1800 seconds/30 minutes)
• uavBatteryFailProb: Probability of a UAV having reduced battery life (default: 0.1)
• initialUavSpacing: Initial spacing between UAVs (default: 50.0 meters)
• uavCrashRadius: Minimum distance before UAVs crash (default: 15.0 meters)
• drawCloud, drawUavs, drawAI, drawAIExtra: Visualization options
• cloudSelection: Choose between different cloud models (1 or 2)

Analyzing Results

After running a simulation, you can analyze the results using the visualization tool:

1. Make sure you've just completed a simulation (variables must be in memory)
2. Run:

   visualize_results

This will generate:

• UAV position plot with operational status
• Battery life distribution histogram
• UAV states pie chart
• Operational status summary
• Detailed statistics in the command window

AI Strategy

The UAV swarm uses a distributed AI approach where each UAV:

1. Takes sensor readings to detect pollutant concentration
2. Communicates its position and findings to nearby UAVs
3. Leaves virtual "pheromones" to mark searched areas
4. Makes decisions based on local information and pheromone trails
5. Avoids collisions with other UAVs
6. Returns home when battery is low

Performance Metrics

After running the simulation, the system reports:

• Number of UAVs that failed to return to their launch spot
• Number of UAVs that crashed into other UAVs
• Number of UAVs that left the bounds of the map