Integrating qualitative and quantitative reasoning spatio-temporal models and their application to robotics TIC2003-07182

Autonomous navigation of real robots includes four problems: mapping, locating, planning and motion control. The mapping problem is been currently solved by representing the environment of the robot by hybrid information: imprecise quantitative measures provided by sensors and the corresponding qualitative representation, mainly by cognitive maps. The rest of the problems in autonomous navigations is been currently solved using quantitative techniques, usually probabilistic, in order to overcome with the imprecision. The idea of an autonomous robot navigating by using qualitative or common sense reasoning comes from the last 80th (i.e. the approaches developed by Kuipers et al.), although they have never been applied to real robots because they assume unreal perfect sensor information. In the 90th decade, all around Europe, a lot of theoretical qualitative models of different aspects of space and time were developed. Although the idea of applying them to robot navigation was very attractive, and some results were obtained in representing qualitatively the environment, no qualitative reasoning process has been used for solving any of the navigation problems on real robots. In this project we want to use qualitative representation enhanced with quantitative imprecise data in a hybrid representation, and qualitative or common sense reasoning to solve the four problems of autonomous real robot navigation. Moreover, we will continue developing theoretical qualitative models of those aspects of space, time and movement that we need in order to solve particular problems. We also are going to deal with sensor integration: sonar, laser and vision based on biological models.