Instantaneous fuzzy modelling for obstacle avoidance

A wheeled mobile robot in its work space has to cope with moving or static obstacles. In order to achieve specific tasks it has to manage with its environment. Two extremes cases can be presented, in one hand when robot has a complete modelling of its workspace, and in an other hand, when there is no a priori knowledge about the external world. First case can be solved with the definition of free paths in the global plan but only if the obstacles configurations are perfectly known. When asynchronous events occur, the system needs a local perception to correct the initial path. In second case robot motion is directly controlled by the local perception and according to the local plan modifications. Robot is so able to manage with statics or dynamics obstacles. Such algorithms are reflex loops. We propose an avoiding strategy based on reflex loop method. This approach is a qualitative one lying on fuzzy reasoning. We develop a method integrating obstacles space configuration and admissible displacements of the system. Obstacle avoidance results from a qualitative local model of the external world, through the estimation of a constraint weight of the current robot configuration in the local environment. We will describe all steps of our reasoning to lead to a robust avoiding strategy in front flexibility and local minima. Example in unknown environment will illustrate the efficiency of our approach