Knowledge Representation and Decision Making for Mobile Robots
Elena R. Messina, Stephen B. Balakirsky
Knowledge is central to a mobile robot's ability to carry out its missions and adapt to changes in the environment it is traversing. The knowledge subsystem must support acquisition of information from external sources, maintain prior knowledge, infer new knowledge from the knowledge that has been captured and provide appropriate input to the planning subsystem. In order to carry out these responsibilities, there are different categories of knowledge required: task (also known as functional or procedural), and declarative, which includes spatial (or metrical). Representation schemes for the various types of knowledge must be chosen so as to provide the best performance and reliability. Many design decisions must be made, taking into account the real-time requirements of the robot control system, the resolution of the sensors, as well as the onboard processing and memory.Decision-making must be tightly coupled with knowledge representation because the decisions must be based on the knowledge available to the robot. Roboticists have drawn from fields as varied as symbolic AI (e.g., state-space search), Operations Research (e.g., cost-benefit-analysis), Economics (e.g., markets and bidding), and Political Science (e.g., voting methods) for inspiration, as well as creating many ad-hoc methods such as behavior-fusion.