A Scenario for Planning Visual Navigation of a Mobile Robot

Abstract

We can classify visual navigation of a mobile robot into two different scenarios. One scenario is given by the world that is out of the scope of our sensing capabilities. We call this world the dark world. In general the robot's goal will be in the dark world. In order to do better than a simple random walk, the robot needs some kind of internal representation that allows it to calculate a long-term plan. On other hand, we have our local position, where we can perceive the world directly through our perception capabilities. We call this scenario the bright world. The main issue about the bright world is that any event in it can have an immediate impact on the performance of the robot. In order to deal with the intrinsic uncertainty and dynamic of the bright world we need appropriate reactive behaviors. These behaviors should be able to couple sensing and action in a synchronized way, in order to allow safe and intelligent navigation. We believe that one of the main challenges of planning for a mobile robot is to develop suitable techniques to deal with the constraints and requirements of each scenario. These techniques should be combined to generate a consistent plan. In this extended abstract we present some of the ideas that we are currently pursuing for the visual navigation of our robots. First we present some ideas about a possible approach for motion planning. Then we support these ideas by analogies with biological systems, and a simplified implementation on a real robot.