An integrated experience-based approach to navigational path planning for autonomous mobile robots

An alternative approach to navigational path planning that integrates an experience-based method with the traditional model-based method is presented. Core issues in using experience-based methods include the content, representation and indexing of past cases, organization of the case memory, retrieval from memory of cases relevant to the current problem, adaptation of retrieved cases to meet the specification of the current problem, and verification of the adapted solution to the problem. It is hypothesized that spatial models of navigation spaces can provide answers to some of these issues in experience-base path planning. The Router system presented examines this hypothesis in the context of path planning in geographical spaces, using a hierarchically organized spatial model of the navigation space to index the cases and to organize the case memory. It also uses model-based reasoning to adapt past path plans and to verify new ones. In addition, Router uses a flexible control architecture that allows for opportunistic selection and integration of the experience-based and model-based methods