Fuzzy Mediation for Online Learning in Autonomous Agents

Abstract

Information fusion is a field that deals with merging multiple inputs to create a single output. Several fusion mechanisms are in place, but none can be easily adapted to accommodate for online learning using a simulator. This paper analyzes the concept of fuzzy mediation as a novel approach to the creation of a framework through simulations of controllers interacting with a controllable object. We propose this new algorithm as a mean to improved methods of learning through positive reinforcement, observation, and direct interaction with the object to be controlled.