Combination of model-based and reactive methods in autonomous navigation

The paper presents three contributions to autonomous navigation of mobile robots: (1) a purely reactive method based on the potential field theory enhanced with a novel procedure to avoid local minima; (2) a topological map building method based on the sensory gradient concept that combined with the reactive module constitutes a hybrid navigation system and (3) another navigation system that uses the robot's behaviors to automatically build a complete navigation model of the environment. All these schemes have been implemented on a Nomad-200 platform and fully tested in indoors environments.