Goal-directed behavior is a hallmark of intelligence. While the majority of artificial intelligence research assumes goals are static and externally provided, many real-world applications involve unanticipated changes in the environment that may require changes to the goals themselves. Goal reasoning, which emphasizes the explicit representation of goals, their automatic formulation and dynamic management, is considered an important aspect of high-level autonomy. Building from these three basic requirements, we describe and apply a framework for surveying research related to goal reasoning that focuses on triggers and methods for goal formulation and goal management. We also summarize current research and highlight potential areas of future work.
展开▼
