Multi-agent Software Control System with Hybrid Intelligence for Ubiquitous Intelligent Environments

Abstract

In this paper, a novel ubiquitous intelligent environment platform and its multi-agent control system are presented. The platform named Distributed Embedded Intelligence Room (DEIR) has been constructed with embedded sensors, actuators and computing devices. All the devices are interconnected using five different physical networks. This platform aims to facilitate realistic data collection and online system performance evaluation. The multi-agent control system incorporating two machine learning algorithms, fuzzy inference and decision tree, has been designed to conform to DEIR architecture. Devices to be controlled are classified based on their possible output states and modelled separately by fuzzy inference agents and decision tree agents in the system. The multi-agent control system with hybrid intelligence shows 11% improvements on overall control accuracy and 84% improvements on learning time compared to its predecessor control system. The vast improvement on computational time shows suitability of the approach towards real-time, embedded applications.