A Multiple Level MIMO FL Based Intelligence For Multi Agent Robot System

ABSTRACT Fuzzy Logic is a many valued logic. It is quite similar to human reasoning, using approximate measures rather than exact. This makes it suitable for linguistic variable analysis. It can be applied to many applications in AI, control and robotics. In this application we discuss the development of an AI system using FL for a dynamic multiple agent robot system. The system has two gaming teams with four robots each with multiple identity assignment. Each robot will have its distinct behavior. To design pure FL based AI, we use fuzzy logic blocks in parallel and series combinations. Further, there is multiple input - multiple output (MIMO) implementation in fuzzy logic blocks. This is necessary to utilize pure fuzzy logic control in AI. The concept was taken from FIRA Micro-Robot World Soccer Tournament (MiroSot). In our setup, there are four robots in each team to be assigned with three different identities; the Forward, two Back and the Goalie. Identity assignment depends on the position of each robot in the play area. Isolation is done to tune each fuzzy logic block individually. Some tuning is performed in a simulator while most in the actual platform. The linguistic approach of FL along with human reasoning nature made it possible to achieve the purpose. Overall, the proposed AI produced expected response experimentally.