Co-Exploration and Monitoring in the Bandwidth Limited Environments
We propose a novel system for co-robotic exploration to aid in the discovery and mapping of scientific phenomena in extreme environments, such as the deep sea, where autonomous robots operate under strong communication bottlenecks. The proposed robotic system learns an unsupervised probabilistic scene model of the environment and then uses this scene model to communicate the spatial distribution of various high-level semantic scene constructs, representing coral species or other substrate types, with a human scientist partner. Such a scene model can enable high-level bi-directional interaction between the human scientist and autonomous robot. We demonstrate the suitability of this approach for co-robotic exploration in low bandwidth environments and quantify how the free parameters of the unsupervised scene model impact the scientific utility of and bandwidth required to communicate the resulting scene model.