Multi-Robot Feedback-Driven Ergodic Coverage Adapts to Unknown Environments

Traditional ergodic coverage methods rely on a pre-specified target distribution of environmental information, which is often unavailable in unknown environments. To address this, Thales Costa Silva and Nora Ayanian introduce a feedback-driven approach that continuously updates a parametric environmental model based on real-time sensor data. Robots then dynamically adjust their trajectories to align their time-averaged spatial distribution with the evolving model, effectively prioritizing areas of high informational value. This allows teams of robots to explore and cover unknown terrains without prior maps, adapting to static or slowly changing conditions.

The framework synthesizes individual control policies for each robot, balancing exploration and exploitation. The authors validate their method through simulations, demonstrating significant improvements in coverage efficiency and resource allocation compared to baseline ergodic techniques. This work has practical implications for applications such as search and rescue, environmental monitoring, and autonomous exploration, where teams of robots must efficiently gather data in unstructured environments without human intervention.