CONTINUOUS CASE-BASED REASONING

Case-based reasoning systems have traditionally been used to perform h igh-level reasoning in problem domains that can be adequately describe d using discrete, symbolic representations. However, many real-world p roblem domains, such as autonomous robotic navigation, are better char acterized using continuous representations. Such problem domains also require continuous performance, such as on-line sensorimotor interacti on with the environment, and continuous adaptation and learning during the performance task. This article introduces a new method for contin uous case-based reasoning, and discusses its application to the dynami c selection, modification, and acquisition of robot behaviors in an au tonomous navigation system, SINS (self-improving navigation system). T he computer program and the underlying method are systematically evalu ated through statistical analysis of results from several empirical st udies. The article concludes with a general discussion of case-based r easoning issues addressed by this research.