Embedding Distributed Learning Algorithms in Wireless Ad-Hoc Control Networks

With the advances in soft computing techniques and agent technologies, the concept of home ambient intelligence is becoming more and more realistic. Living in a building that adapts itself to the users and assists them in reducing their energy consumption is now within reach. The main technical barrier comes from hardware: servers and industrial control networks do not fit in a house. With the availability of dedicated wireless solutions and low-cost small computation units, the platform to implement task distribution in a control network is now feasible and cost efficient. This paper explores the possibilities of fitting a distributed learning algorithm for home ambient intelligence in a wireless network of sensors and actuators, driven by very limited microcontrollers. The chosen hardware platform is the WACNet: Wireless Ad-hoc Control Network. The concept of WACNet is introduced and the test-bed developed for its study is explained. The fuzzy learning algorithm is then introduced and its implementation is studied. The results of a test are provided and some conclusions are drawn, mainly focusing on accuracy and the algorithm’s response to different rule selection criterions.