In the design of autonomous mobile robots, databases have been used mainly to store information on the environment in which the device is to operate. For most of the models and ready systems, the database when used, is not a stand alone component in the system, rather it is only intended to keep static information on the disposition and properties of objects on the map.
2In this thesis is implemented an intelligent database. This database is called intelligent because it is knowledge-based. It combines static facts to build more information. An intelligent database such as this one will be a plus for an intended autonomous machine such as the PSUBOT wheelchair being developed in the Department of Electrical Engineering. The database will make intelligent decisions as an intelligent function of the the central control module of the system (i.e, find a global optimum path, recognize details in the building, support sensory integration). The database will also serve as the core of pattern recognition and localization of the wheelchair inside the building. The database has been implemented on a PC-386SX with 640K base memory and a clock resolution of 20MHz. It is composed of a relational database, a knowledge base, a set of management routines and programs to perform tasks such as global path planning and image matching for localization. The relational database has been implemented using the BORLAND PARADOX3 database management system. The knowledge base part of 3 the database has been implemented in a combination of BORLAND Turbo Prolog (for intelligent tasks) and BORLAND Turbo C++ (for tasks requiring faster computation power). Global path planning has been implemented with a knowledge-based approach rather than a conventional graph method in order to match the hierarchical description of a building and to add more intelligence power. Localization is the key problem for which we use image matching. The wheelchair needs to possess the capability to recognize where it is in the building. Therefore, in order to perform this task, the matching of the current image of the scene with template images of candidate locations is performed. The image matching has been implemented to match two images described each as a set of straight lines. The matching method is correspondence matching between the two sets of features with the criteria being the best acceptable match. Image matching is used for localization of the wheelchair inside the building based on matching of the currently perceived image with template images of locations previously stored in memory.