Reasonable Machines: Analogical Reasoning in Autonomous Agent Design

Abstract

AbstractThis paper focuses on agents for open environments (AOE) and on the use of analogical reasoning (AR) in their design. The autonomous mobile robot is an example of an AOE. It operates in an open, dynamic environment, i.e. unpredictable but familiar. An AOE should be a reasonable machine: it should act as we would for the same environment, history and goals. A reasonable machine has responses that are mediated among its goals and actions in the current context. AR can be used to design reasonable machines that implement this mediation. An agent using AR reasons from known examples to define actions in the current context. This paper proposes a form of AR called Analog Logic and discusses its potential to achieve AOE design goals. AR provides philosophical grounding of the symbols used and a new approach for dealing with continuous symbol systems and agent communication.KeywordsMembership FunctionFuzzy LogicMobile RobotOpen EnvironmentAutonomous AgentThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.