You are here: Home > For Researchers > Projects > Risk-averse Control and Dynamic Optimization Bridging the Gap Between Robust and Stochastic Control

Risk-averse Control and Dynamic Optimization Bridging the Gap Between Robust and Stochastic Control

From 01-01-2018 to 31-12-2021

Description

Ubiquitous and inevitable as it is, uncertainty needs to be taken into account when we make decisions. In engineering, automated decision making is referred to as control. There exist two tools in an engineer's toolbox when it comes to designing a control system which is resilient to uncertainty. The first is the so-called robust approach where we try to protect the system against worst-case events which, nevertheless, are unlikely to happen in practice. The adoption of such a pessimistic strategy leads to poor performance. The second approach is known as stochastic, where we opt for policies which lead to optimal average performance. However, the impossibility of fully capturing the underlying distribution creates a yawning chasm between theory - with all its convenient assumptions - and practice. These two decision making practices have been being studied separately with no obvious connection to another and, most importantly, without any halfway choices. In this project we will propose a new decision making paradigm which bridges the gap between the robust and stochastic approaches using the concept of risk. This will furnish the designer with a new powerful tool and will enable the operation of systems in highly uncertain contexts. The key contributions of this project will be (i) an arsenal of theoretical analysis and design tools, (ii) computational methods and software to facilitate the technological uptake of our developments (iii) demonstration on a bioreactor.