HYSTE: A HYbrid System for Thorough Exploration

This paper describes HYSTE, a system for thorough exploration of unknown environments by autonomous mobile robots. The system is based on a hybrid reactive/deliberative architecture that adds a middle layer for mapping and localization [11]. The reactive layer is implemented using a behavior-based motor schema approach. Motor schemata are parametrized and grouped into distinct behaviors, each of which causes slight but noticeable changes in navigation. A middle layer accumulates sensor information into an evidence grid and a topological connectivity graph. These maps are used by the deliberative layer to set goal points in unknown space to be achieved by the reactive layer. The deliberative layer can also select a behavior for the reactive layer, escalating the aggressiveness of the behavior selected if attainment of a goal point may require movement through dense obstacles or small gaps. Each layer is parametrized to allow for a maximum amount of durability and adaptiveness in real world environments. Results from several scenarios display the fully functional performance of HYSTE.