A Particle Swarm Optimization-Based Cooperation Method for Multiple-Target Search by Swarm UAVs in Unknown Environments

Paper Link : https://www.researchgate.net/publication/350149777_A_Particle_Swarm_Optimization-Based_Cooperation_Method_for_Multiple-Target_Search_by_Swarm_UAVs_in_Unknown_Environments

Overview

This repository contains a swarm UAV coordination system designed for multi-target search, optimized payload drops, and inter-UAV collision avoidance. The system is based on a Multi-Target Particle Swarm Optimization (MTPSO) approach, inspired by Particle Swarm Optimization (PSO), to enable decentralized cooperative UAV control.

The swarm UAVs utilize onboard sensors and real-time data sharing to dynamically adjust flight paths, detect targets, and deploy payloads efficiently in unknown environments.

Key Features

• Decentralized Swarm Coordination: UAVs operate in a networked environment, sharing limited data to improve efficiency.
• Multi-Target Search Optimization: Implements an adaptive PSO variant (MTPSO) for improved search performance.
• Inter-UAV Collision Avoidance: Ensures safe flight paths using dynamic trajectory adjustments.
• Optimized Payload Delivery: UAVs determine the best drop points in real-time.
• Simulation Integration: Uses Ardupilot’s Simulation in the Loop (SITL) for realistic testing.
• Human Detection: Vision-based processing using Deep Learning models for identifying targets.
• Ground Control Station (GCS) Interface: A control interface to manage UAVs and monitor their status.

---

Directory Structure

The repository consists of the following major components:

gcs_client4/

Contains scripts related to the Ground Control Station (GCS), which communicates with UAVs.

• gcs_client4.py - Manages communication between the GCS and UAV swarm.

new_on_uav/

Core UAV control scripts, including swarm behavior, emergency handling, and network management.

Key Scripts:

• Swarm2.py - Implements swarm behavior and MOPSO-based navigation.
• SwarmBot.py - Defines individual UAV control, including navigation and state tracking.
• Uav_to_GCS.py - Manages data exchange between UAVs and the GCS.
• helper.py - Contains utility functions for waypoint calculations, distance measurement, and velocity adjustments.
• emergency_landing.py - Handles emergency UAV landings.
• back_to_adhoc.py - Ensures UAVs remain connected in Ad-Hoc networking mode.
• modified_server8.py - Alternative UAV server implementation.

Supporting Files:

• wp_list, number_of_UAVs, weight_matrix - Configuration files for swarm parameters and navigation.
• mav.tlog, eeprom.bin - Log files for debugging and data recording.

VISION/

Includes scripts and data for human detection and vision-based navigation.

Key Scripts:

• human_detection.py - Uses deep learning to identify humans in UAV camera feeds.
• VideoGrabber.py - Handles real-time video streaming from UAV cameras.
• LocalPlanner.py - Computes local navigation paths based on detected objects.
• calc_gps.py, gps_correction.py - GPS processing utilities.

Deep Learning Model Files:

• darknet/files/human_detection.cfg, human.names, tiny_human_detection.cfg - Configuration for object detection models.

GUI/

The Ground Control Station graphical interface.

Key Scripts:

• GUI.py - Main GUI application for monitoring and controlling UAVs.
• working_gui.py, new_gui.py, tested_gui.py - Various versions of the control interface.
• connectiontouavs.py - Manages network connectivity between the GUI and UAVs.
• LAND_ALL_UAVS.py - Emergency landing control for all UAVs.

---

Installation and Setup

Prerequisites

Ensure you have the required dependencies installed:

pip install -r /requirements.txt

Running the System

1. Start the GCS Interface:

   python3 CORE_GUI_CODE/GUI.py

2. Launch Swarm Control:

   python3 CORE_UAV_CODE/Swarm2.py

3. Monitor Video Feeds:

   python3 CORE_UAV_CODE/VISION/VideoGrabber.py

4. Emergency Landing (if required):

   python3 CORE_UAV_CODDE/emergency_landing.py

---

Algorithmic Approach

Multi-Target Particle Swarm Optimization (MTPSO)

The UAV swarm utilizes a modified PSO-based control mechanism:

• Each UAV adjusts its position based on personal best, global best, and inter-UAV collision avoidance factors.
• UAVs dynamically share target data to optimize the search process.
• Payload drops are assigned based on proximity and availability.
• The system integrates various inertial functions to balance exploration and exploitation.

Collision Avoidance

• Ensures UAVs maintain safe distances using consensus-based adjustments.
• The Dsafe parameter prevents UAV clustering and reduces congestion.

Simulation and Testing

• The system is tested using Ardupilot SITL.
• Multiple hyperparameters such as FOV, velocity, and swarm size are optimized to achieve efficient results.

---

Future Improvements

• Adaptive UAV Formation: Dynamic restructuring of UAVs based on detected targets.
• Enhanced Target Prediction: Using reinforcement learning for improved target identification.
• Energy-Aware Navigation: Optimizing routes to extend UAV battery life.

---

Contributors

• Parth Mahajan
• Aniket Gupta
• Aman Virmani

For further details, refer to the A Particle Swarm Optimization-Based Cooperation Method for Multiple-Target Search by Swarm UAVs in Unknown Environments paper.

---

License

This project is open-source and available under the MIT License.

##To Cite Us Citation : Gupta, Aniket & Virmani, Aman & Mahajan, Parth & Nallanthigal, Raghava. (2021). A Particle Swarm Optimization-Based Cooperation Method for Multiple-Target Search by Swarm UAVs in Unknown Environments. 95-100. 10.1109/ICARA51699.2021.9376529.