This repository contains the implementation of a multi-robot cooperative tracking system developed for the Intelligent Distributed Systems course.
The goal is to enable a team of mobile robots (LIMO) to collaboratively localize and track a moving human agent in a shared environment while maintaining a coordinated formation around the target. The system combines decentralized estimation, vision-based perception, and model-based control.
The system consists of three LIMO mobile robots that cooperatively track a moving person while maintaining a coordinated formation around the target. The robots combine vision-based perception, cooperative state estimation, and model-based control to ensure consistent relative positioning with respect to the human agent.
Each robot operates with local sensing and contributes to a shared estimation process through inter-agent measurements.
- Implemented an Interacting Multiple Model / decentralized EKF-based cooperative localization framework
- Enables inter-robot relative measurements
- Fusion of:
- wheel odometry
- stereo camera measurements
- relative observations between agents
This allows consistent multi-agent localization even under partial observability.
- Marker-based detection for inter-robot and robot-person observations
- Multi-source visual fusion for robust target identification
- Model Predictive Control (MPC) for trajectory tracking and formation maintenance
- Each robot computes local control actions to:
- track the moving target
- preserve formation constraints relative to other agents
- ensure smooth coordination in dynamic environments
Each LIMO robot runs a local pipeline:
-
Perception:
- Marker detection
-
State Estimation:
- Cooperative EKF with inter-agent measurements
- Fusion of odometry + stereo vision + relative observations
-
Control:
- MPC-based motion controller
- local trajectory generation for tracking and formation keeping
Communication between agents is handled using ROS2 topics.
- Python
- ROS2
- OpenCV
- NumPy
- Nonlinear estimation (EKF)
- Model Predictive Control (MPC)
- Stereo vision
- Decentralized multi-robot state estimation
- Cooperative localization using inter-agent measurements
- Marker-based robot-to-robot detection
- MPC-based control in dynamic environments
- Formation-based multi-robot tracking of a moving human
- Simulation-first → real robot deployment pipeline
- Maintain a dynamic formation of robots around a moving human target
- Achieve robust multi-robot tracking of a moving human
- Maintain consistent localization under uncertainty
- Leverage inter-robot cooperation to improve estimation accuracy
- Transition from simulation to real-world robotic deployment
Marco Misseroni and Federico Battisti
MSc Mechatronics Engineering – Electronics and Robotics
University of Trento