Multi-robot Collaborative Simultaneous Localization and Mapping

Advancements in multi-robot systems represent a significant frontier in robotics, with growing applications in domains such as agriculture, search and rescue, industry, domestic environments, and transportation. A key challenge in these settings is enabling multiple robots to operate collaboratively, safely, and efficiently while navigating complex, dynamic environments. This thesis focuses on the development of a multi-robot collaborative SLAM (Simultaneous Localization and Mapping) system designed to provide robust and accurate localization while enabling the coordinated and efficient exploration and mapping of unknown environments. By employing a centralized framework, the system integrates sensor data and robot dynamics to maintain global consistency across the multi-robot network. Factor graph formulations are employed as a powerful tool to jointly solve SLAM, decision-making, and trajectory planning for the multi-robot system, implemented using the GTSAM library. Within the SLAM component, the framework optimizes the entire system’s state by integrating motion constraints and sensor measurements while simultaneously building the map and exploring frontiers to discover unknown areas. For decision-making, discrete factor graphs are used to assign frontiers to individual robots, ensuring an efficient allocation of exploration tasks. In local trajectory planning, the factor graph solution naturally extends to jointly optimize robot paths, promoting efficient, safe navigation while enforcing collision avoidance and coordinated behaviors across the team. Experimental evaluations are conducted in simulation and indoor environments using a team of TurtleBot3 robots, with a Vicon motion capture system providing ground truth references. These experiments assess mapping accuracy, localization reliability, and efficiency in multi-robot exploration. The results show the effectiveness of factor graph-based frameworks enhancing multi-robot collaboration in complex, unstructured environments, while also pointing out some limitations and directions for further improvement.