Collaborative ADMM-based SLAM algorithm for fleets of mobile robots in a ROS 2 - Webots environment

This thesis addresses the challenges of Simultaneous Localization and Mapping (SLAM) in mobile robotics, where a single agent typically estimates its position while constructing a map. Traditional Pose Graph Optimization (PGO) methods suffer from accumulated noise, scalability issues, and robustness concerns due to sensor drift and failures. For this reasons Collaborative Pose Graph Optimization (C-PGO) was introduced, allowing multiple agents to work together, introducing inter-agent loop closures that improve scalability, and robustness through shared observations and information exchange. This project focuses on the practical implementation, through simulation in a ROS 2 – Webots environment, of the Incremental Manifold Edge-based Separable ADMM (iMESA) algorithm, which enables real-time state estimation for multi-robot teams while relying on sparse communication. This thesis develops methods for extracting odometry and pose measurements from simulation, implements a decentralized communication strategy for detecting shared landmarks, and integrates relative pose estimation to enhance localization accuracy. By testing iMESA in a realistic setting, this thesis demonstrates its scalability and robustness for multi-robot SLAM.