A Supervisor Αgent-Based on the Markovian Decision Process Framework to Optimize the Behavior of a Highly Automated System

In this paper, we explore how MDP can be used as the framework to design and develop an Intelligent Decision Support System/Recommender System, in order to extend human perception and overcome human senses limitations (because covered by the ADS), by augmenting human cognition, emphasizing human judgement and intuition, as well as supporting him/her to take the proper decision in the right terms and time. Moreover, we develop Human-Machine Interaction (HMI) strategies able to make “transparent” the decision-making/recommendation process. This is strongly needed, since the adoption of partial automated systems is not only connected to the effectiveness of the decision and control processes, but also relies on how these processes are communicated and “explained” to the human driver, in order to achieve his/her trust.