Loris Di Natale

Dr.

Loris Di Natale

PhD Student

Alumni

Empa, EPF Lausanne

loris.dinatale@empa.ch

Curriculum Vitae

Automation is everywhere and more and more data is available - Machine Learning can help bring both together

Loris was a PhD student on data-driven control of energy systems in the Urban Energy Systems lab at Empa. He was supervised by Prof. Colin Jones in the Automatic Lab at EPFL. He received his Bachelor's in Mathematics (2017) and his Master's in Energy Management and Sustainability (2019) from EPFL.

After completing his PhD Loris took up a position at Zurich Soft Robotics.

Scientific Publications

Article

The Potential of Vehicle-to-Grid to Support the Energy Transition: A Case Study on Switzerland

Loris Di Natale

Luca Funk

Martin Rüdisüli

Bratislav Svetozarevic

Philipp Heer

Giovanni Sansavini

Volume

Number

Article

Deep Reinforcement Learning for Room Temperature Control: A Black-box Pipeline from Data to Policies

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

Inproceedings

Theoretical Foundations

Energy systems

Near-optimal Deep Reinforcement Learning Policies from Data for Zone Temperature Control

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

Booktitle

2022 IEEE 17th International Conference on Control \\& Automation (ICCA)

Pages

698-703

Inproceedings

Theoretical Foundations

Computational Methods

Energy systems

Lessons Learned from Data-Driven Building Control Experiments: Contrasting Gaussian Process-based MPC, Bilevel DeePC, and Deep Reinforcement Learning

Loris Di Natale

Yingzhao Lian

Jicheng Shi

Colin Jones

Booktitle

2022 IEEE 61st Conference on Decision and Control (CDC)

Pages

1111-1117

Miscellaneous

Energy systems

Physically Consistent Neural Networks for building thermal modeling: theory and analysis

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

Miscellaneous

Theoretical Foundations

Uncertainty Propagation via Optimal Transport Ambiguity Sets

Loris Di Natale

Inproceedings

Scalable and correct by design synthesis

Computational Methods

Computationally Efficient Reinforcement Learning: Targeted Exploration leveraging Simple Rules

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

Booktitle

Proceedings of the 2023 62nd IEEE Conference on Decision and Control (CDC)

Author Address

Singapore, Singapore

2023

Miscellaneous

Scalable and correct by design synthesis

Computational Methods

Computationally Efficient DRL

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

2023

Article

Control in a data rich world

Theoretical Foundations

Data-driven adaptive building thermal controller tuning with constraints: A primal–dual contextual Bayesian optimization approach

Wenjie Xu

Bratislav Svetozarevic

Loris Di Natale

Philipp Heer

Colin Jones

Journal

Applied Energy

Volume

358

Number

122493

2024

Inproceedings

Scalable and correct by design synthesis

Computational Methods

Computationally Efficient Reinforcement Learning: Targeted Exploration leveraging Simple Rules

Loris Di Natale

Bratislav Svetozarevic

Philipp Heer

Colin Jones

Booktitle

Proceedings of the 2023 62nd IEEE Conference on Decision and Control (CDC)

Author Address

Singapore, Singapore

2023

Miscellaneous

Control in a data rich world

Theoretical Foundations

Data-driven adaptive building thermal controller tuning with constraints: A primal–dual contextual Bayesian optimization approach

Wenjie Xu

Bratislav Svetozarevic

Loris Di Natale

Philipp Heer

Colin Jones

2024

News & Insights

27-03-2023

Going back to what we know: Injecting physical insights into neural networks

Researcher blogs

While neural networks often achieve impressive performance, they sometimes fail to capture simple concepts that humans understand intuitively – a gap that may even put lives at risk. But can we transfer our human knowledge to neural networks to help them learn more meaningful solutions? NCCR Automation researcher Loris Di Natale has a plan to make sure automated machines don’t freeze us out.

Read more about Going back to what we know: Injecting physical insights into neural networks