Modelling Controllers for Cyber Physical Systems Using Neural Networks

TL;DR

This paper explores using neural networks to approximate Model Predictive Controllers in cyber-physical systems, aiming to enable faster, real-time control by addressing data and learning challenges.

Contribution

It investigates various imitation learning methods for neural network control of cyber-physical systems and evaluates their advantages and limitations.

Findings

Neural networks can effectively approximate MPC functions.

Imitation learning methods vary in performance and data requirements.

Challenges include non-i.i.d data and stability issues.

Abstract

Model Predictive Controllers (MPC) are widely used for controlling cyber-physical systems. It is an iterative process of optimizing the prediction of the future states of a robot over a fixed time horizon. MPCs are effective in practice, but because they are computationally expensive and slow, they are not well suited for use in real-time applications. Overcoming the flaw can be accomplished by approximating an MPC's functionality. Neural networks are very good function approximators and are faster compared to an MPC. It can be challenging to apply neural networks to control-based applications since the data does not match the i.i.d assumption. This study investigates various imitation learning methods for using a neural network in a control-based environment and evaluates their benefits and shortcomings.