Functional multi-objective optimization under uncertainty
(another grant application)
I have just submitted a grant application to the Swedish Science Council with the title "Functional multi-objective optimization under uncertainty". If granted it would pay for a PhD student + supervision time.
The topic is an update on the series of application I have submitted over the last few years - most recently FORMAS 2024-10: Theories and tools for climate policy exploration (rejected a few weeks ago) - with a bit more computer science and physics.
Abstract
A crucial part of many engineering and societal decisions is to find the right trade-off between multiple conflicting objectives. In general, this is known as a multi-objective control problem. There is a system, a space of parameters or control policies, and objectives to be met. No control policy is best, but some are Pareto-optimal in that one can not improve one objective without worsening another. The goal of multi-objective control is to identify optimal regions in the control space, and to understand which trade-offs the optimal control policies make.
The goal of the project is to develop a system of generic (adaptable) and verified (trustable) open-source software for exploring safe and Pareto optimal solutions of multi-objective optimal control problems in collaboration with domain experts from physics and sustainability science.
We will explore three core questions: how to 1) encode stylized decision processes? 2) encode preferences and goals? 3) explore alternative policies? This will be done in the context of multi-objective control problems under uncertainty, using domain-specific languages for exploring their solutions.
The underlying motivation is the profound transformation of our society needed to tackle climate change and the sustainable energy transition. In a nutshell: a transformation towards a fossil fuel free society must itself be sustainable. The project will contribute to ensuring the decision process can be trusted and verified.
The topic is an update on the series of application I have submitted over the last few years - most recently FORMAS 2024-10: Theories and tools for climate policy exploration (rejected a few weeks ago) - with a bit more computer science and physics.
Abstract
A crucial part of many engineering and societal decisions is to find the right trade-off between multiple conflicting objectives. In general, this is known as a multi-objective control problem. There is a system, a space of parameters or control policies, and objectives to be met. No control policy is best, but some are Pareto-optimal in that one can not improve one objective without worsening another. The goal of multi-objective control is to identify optimal regions in the control space, and to understand which trade-offs the optimal control policies make.
The goal of the project is to develop a system of generic (adaptable) and verified (trustable) open-source software for exploring safe and Pareto optimal solutions of multi-objective optimal control problems in collaboration with domain experts from physics and sustainability science.
We will explore three core questions: how to 1) encode stylized decision processes? 2) encode preferences and goals? 3) explore alternative policies? This will be done in the context of multi-objective control problems under uncertainty, using domain-specific languages for exploring their solutions.
The underlying motivation is the profound transformation of our society needed to tackle climate change and the sustainable energy transition. In a nutshell: a transformation towards a fossil fuel free society must itself be sustainable. The project will contribute to ensuring the decision process can be trusted and verified.