Cognitive Collision Prediction & Path Planning Of Mobile Robots: A Qualitative Approach

Abstract

means of nonmonotonic temporal logic and qualitative reasoning techniques. In order t o meet the on l ine limitations, t h e method of discrete i terat ion is exploited t o produce t h e path without running the goal seeking algorithm again. The above techniques a r e implemented in a program. The paper demonstrates our approach through an example. 1. I N T R O D U C T I O N Collision prediction, coll ision avoidance and path planning problems can all be formulated in a similar way, they all begin with a given present state and some assertions on t h e goal in a defined scenario, and t h e robots a r e required t o predict t h e possible collision and decide upon the control t o achieve t h e goal o r subgoals satisfactorily. The abstract knowledge in all cases along with the spatiotemporal coordination remain the same. In t h e path planning case the concentration is on the spat ia l argument and in coll ision prediction avoidance t h e temporality is mainly concerned. Collision avoidance makes use of both. 1 1 . T h e p r o b l e m In ordcr t o sce what a r e the problems if one t r ies t o predict cvcnts in the ruturc and control them, suppose tha t the two dimensional scenario is composed of two mobile robots M I and M P with the preset goals G and G P respectively, approaching cach othcr , and the i r t ra jec tor ies intersect spat ia l ly as in Fig. 1. Each robot has a potent ia l behavior [14] in t h e lack of any o ther obstacle interrupting its t r a j e c t o r y (Fig.1-a). In case of collision only par t s of the potent ia l behavior un t i l the collision point is manifested and a f t e r tha t there is a new behavior which doesn't necessarily terminate t o the previously set goal (Fig.1-b).