Collaborative navigation of heterogeneous robots in unknown environments poses significant challenges due to sensing, communication, and computational limitations. In this work, a lead robot navigates toward a target with the assistance of a mobile sensor robot (e.g., an airborne drone), whose task is to explore the environment and communicate relevant map information under bandwidth constraints. We propose a framework in which the sensor determines both which map regions to transmit and which navigation action to execute online, while also predicting unexplored regions of the environment. To this end, we present $β$-Sparse Gaussian Processes, a variational sparse Gaussian Process (GP) model for task-aware inducing point selection under cardinality constraints. Furthermore, we develop an action-selection strategy that balances task relevance with exploration. Simulations on Mars and Earth maps show that the proposed framework outperforms all considered bandwidth-constrained baselines, reducing accumulated path cost by approximately 35% compared to no communication, while decreasing transmitted information by 94% relative to the full-information GP baseline. The results further highlight the benefits of GP-based map prediction over methods without prediction, as well as the advantages of navigation-aware region selection over approaches that do not prioritize task-relevant regions.
翻译:暂无翻译