Integration of schema-based behaviors and variable-resolution cognitive maps for stable indoor navigation

Presents a situation-driven reconfigurable architecture for a schema-based robot controller which is operating in partially known indoor environments. Instead of the purely repulsive forces from the obstacles, we employ the linear combination of the repulsive and the detouring forces around the obstacle, so that the obstacles generate virtual forces pushing the robot toward the goal rather than just blocking it. By reconfiguring the detouring direction and the activation level of each schema based on the geometric and/or topological information provided by the variable resolution maps, the robot can navigate through the environments in very natural way. Simulation results show that our approach can remarkably improve the intrinsic defects of traditional potential field-bused schemas.