Robot Navigation Using Integrated Retrieval of Behaviors and Routes

摘要

RUPART1 is a hybrid robot control system for navigating a real-world, academic building. Hybrid robot control systems provide robust low-level navigation together with strategic planning abilities. While a hybrid system may produce better overall performance, it is often a complex system that must balance reactive with deliberative tasks, and where learning and adaptation are more difficult to achieve. RUPART ad-dresses the issues of complexity, balance, and learning by us-ing a single case-based-reasoning (CBR) system to store and retrieve cases for both its reactive and deliberative systems. At this stage, the CBR system contains behavior cases, which determine the reactive actions of the robot, and route cases, which determine strategic plans for navigating to a goal lo-cation. RUPART will be extended in the future to use CBR to manage multiple goals and higher-level strategic decision-making. This first-stage RUPART system retrieves and ap-plies cases that are suitable to its needs: behavior cases to control its reactive behaviors, route cases when planning to reach a new goal. The resulting system can learn new route and behaviors, without the complexity of multiple reasoning and learning algorithms usually entailed by a hybrid control system.