Global and local planning in the robotics domain

The top-level software architecture for the mobile robot project, codenamed Turnip 1, is presented. Global planning is the process which generates an ordered sequence of sub-goals which culminate in the achievement of a required task. In Turnip, the world knowledge base has been implemented as a Prolog knowledge base defining topological relationships between symbolic entities (doors and rooms). A depth-first search strategy generates an ordered sequence of landmarks (specific entities) to be reached. To reach each landmark requires both perception and local planning. Perception is required to recognise the landmark and to describe those parts of the world which are filled. Local planning is needed to generate a point-for-point route which takes one from the current location to that landmark. A framework is suggested for providing some useful solutions to the task of local planning