Efficient and Resilient Multi-Robot Exploration in Complex and Unknown Indoor Environments

Multi-robot exploration in complex indoor spaces presents numerous challenges, including high-density layouts, varied environmental conditions, and the need for efficient coordination among agents. This work assesses the performance of twelve exploration strategies, considering factors like initial robot distribution, environmental segmentation, and robot failure conditions. While frontier-based strategies like classic Frontier Closest have demonstrated acceptable performance in many areas, they suffer from limitations, notably oscillatory behaviors and leader-follower conditions specifically in high-density structures. This is the result of its nature, which picks frontiers greedily. To address the issue, another goal selection strategy Unknown Closest is provided. This not only prevents the problem of leader-follower conditions but also minimizes the potential for oscillation. While it is known that frontier-based approaches are not well suited to dense environments, we evaluated the performance of each exploration strategy when they are specifically supported with a task allocation strategy. Distributing robots initially in the environment is also investigated to understand the performance effects of each tested distribution technique. To evaluate the strategies' robustness factor, destructive mine conditions are simulated and they are randomly placed in the different complexity of building floor plans.