A New Approach in Controlling the Compressor of the Vehicle Air Conditioning System

This paper presents a first "proof of concept" for a reinforcement learning based controller used for a simple vehicle air conditioning system (AC) simulation. Our approach aims to control a vehicle aggregate locally using reinforcement learning (RL) techniques and to control the global system composed of these aggregates using multi-agent system (MAS) techniques. This concept allows us to develop control systems for vehicles that are able to adapt to their environment and to learn without any external advisor