Research Project

SWARM Robotics Platform

Distributed Multi-Agent Coordination System

Project Lead β€’ Electronics & Robotics Club, BITS Goa β€’ Jan 2025 - Present

Overview

Leading the development of a custom low-cost motion capture and multi-agent coordination system for swarm robotics research at BITS Goa. This open-source project focuses on making swarm robotics accessible by creating affordable hardware and sophisticated software for drone coordination, collaborative mapping, and distributed intelligence.

Key Innovations

  • πŸ’° Engineered custom motion capture system under $50 (vs commercial $50,000+ systems)
  • 🎯 Developed active and passive IR marker system for unique drone identification
  • πŸ€– Built swarm coordination platform handling multi-drone state estimation
  • πŸ—ΊοΈ Researching novel RGB-based localization using custom ESP32 drones
  • 🌐 Investigating Gaussian Splatting for real-time 3D scene reconstruction

Technical Implementation

Custom Motion Capture System

Salvaged high frame-rate PS3 Eye cameras and IR filters from floppy disks to create an accurate motion capture system at 1/1000th the cost of commercial solutions. Developed custom calibration software for intrinsic and extrinsic camera parameters.

IR Marker System

Created both active (LED-based) and passive (reflective) IR markers for drone tracking and unique identification. The system can track multiple drones simultaneously with sub-centimeter accuracy.

Swarm Algorithms

  • Formation control for coordinated movement patterns
  • Collaborative construction algorithms for building structures
  • Multi-agent SLAM for distributed mapping
  • Both centralized and decentralized control approaches

Novel Research Areas

Exploring RGB-based localization methods as an alternative to traditional IR tracking, and implementing Gaussian Splatting techniques for real-time 3D reconstruction and collaborative mapping by drone swarms.

Technologies Used

ROS 2 Gazebo Computer Vision Python C++ ESP32 Motion Capture SLAM Gaussian Splatting OpenCV