Building Markov Decision Process Based Models of Remote Experimental Setups for State Evaluation

Abstract

Remote Access Laboratories (RAL) are online environments that allows the users to interact with instruments through the Internet. RALs are governed by a Remote Laboratory management system (RLMS) that usually provides the specific control technology and control policies with regards to an experiment and the corresponding hardware. Normally, in a centralized RAL these control strategies and policies are created by the experiment providers in the RLMS. In a distributed Peer-to-Peer RAL scenario, individual users designing their own rigs and are incapable of producing and enforcing the control policies to ensure safe and stable use of the experimental rigs. Thus the experiment controllers in such a scenario have to be smart enough to learn and enforce those policies. This paper discusses a method to create Markov's Decision Process from the user's interactions with the experimental rig and use it to ensure stability as well as support other users by evaluating the current state of the rig in their experimental session.