Purposeful and reactive navigation based on qualitative path planning and fuzzy logic

Abstract

An important problem in robotics research is to develop methods for safe, robust, and efficient navigation. To be able to navigate purposefully in its environment, a robot has, first, to have some a priori knowledge about its environment, so that it can plan the paths necessary for accomplishing its tasks, and, second, to be able to react to any unforeseen obstacles that may arise it its way during the execution of its paths. Since real environments are usually complex and dynamic, the robot cannot have complete and correct a priori knowledge about them. To aggravate the situation, the robot's sensors usually deliver noisy readings, so even its runtime perceptions are not absolutely correct. Towards addressing the problem of purposeful yet reactive navigation without assuming either precise a priori knowledge or noise-free runtime sensory input, we have developed a navigation method integrating a qualitative path planner and a fuzzy controller. This method has been implemented and evaluated in the context of David, a Nomad 200 robot.