Sebastien Gros
Background and activities
Current research
My main current research concerns safe Reinforcement Learning and Data-Driven MPC. I currently have 5 PhD students. This work is strongly connected to the industry, with large industrial partners such as DNV GL and Konsberg Marine.
I also co-supervise several PhD projects related to the Artificial Pancreas problem, multi-rotor wind turbine control. Finally, I supervise and cosupervise several industrial PhD projects based at Volvo (Sweden), focusing on powertrain control, 2nd life of batteries, and traffic management.
To access my publications, please consult:
- https://www.researchgate.net/profile/Sebastien-Gros-2
- https://scholar.google.no/citations?user=38fYqeYAAAAJ&hl=en
MSc thesis / Specialization projects for 2022-2023
I propose several projects all revolving around an existing smart house running in Trondheim. The system is equipped with two heat pumps (4 indoor units), real-time power measurements, multiple climate sensors, and a cutting-edge ventilation system with heat recovery. All equippments are IoT enabled, and the heat pumps are currently operated by an MPC / MHE scheme against the local spot market. This smart house aims at serving as a demonstrator and experimental platform to prepare a wide deployment of smart house solutions in Norway. A number of questions are still under investigation, summarized in the following projects.
Optimal control of battery storage in Smart Houses
With the raising amount of variations in the electricity prices, battery storage is becoming increasingly interesting for individual houses. This is especially the case for houses that have PVs, and if the battery storage is built from recycled EV batteries. The start-up company ChainPro is currently developping this business in Trondheim. In this project, we will investigate the development of advance optimization solutions to size the battery storage and exploit it optimally, in order to minimize the cost of electricity for the user. ChainPro has more than a dozen houses equipped with their system, providing data and an experimental testing ground for the solution we develop.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no),
Advance IoT-based smart house climate control
In a previous MSc project, we have developped a software architecture to manage the smart house from a Raspberry Pi. The software allows the management of connected heat pumps, and the reading and logging of multiple sensors (temperature, power, etc.). In this project, the cutting-edge ventilation system with IoT capabilities will be added in the in the Pi software. We will then investigate the possibility of deploying smart control algorithms for the ventilation system, in order to optimize the climate of the house against the electricity spot market and nettleie. The student in this project will have to familiarise him/herself to the existing software infrastructure, develop it further, and finally start building solutions for the smart management of the ventilation system.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no),
IoT hardware and software for Smart Houses
The smart house is currently operated by a Raspberry Pi. While cheap and effective, it is not obvious that this choice of hardware is the best one. In particular, a light computer with a stronger computational capability may be benefitial. In this project, we will investigate the use of different computer architecture, and develop IoT software for these architecture for data management and control.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no)
Hybrid Digital Twin for smart house control
An early modelling of the smart house dynamics has been performed, using classic System Identification (SYSID) tecniques. A large amount of data has been collected on the house. The data collection is ongoing. In this project, we would like to investigate further modelling approaches for smart houses. The end goal would be to develop a modelling technique that can be deployed on different houses with a limited amount of engineering work. Methods based on Hybrid Digital Twins, where physical modelling is combined with Machine Learning (ML), are considered as the most promising. Various ML techniques will be considered, such as Gaussian Processes and Bayesian Neural Networks.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no),
Learning-based Optimization of Smart Houses
The smart house heating system is currently operated by an advance MPC / MHE scheme. Due to the difficulty of modelling the dynamics of a house accurately, the model on which this control scheme is based is far from perfect. In this project, we will investigate the use of learning techniques in order to compensate for the model inaccuracy, and improve the control performance of the control scheme using the data collected on the house. We will focus on Reinforcement Learning as a learning approach.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no)
Improved MHE / MPC scheme for smart house control
The current smart house is managed by an advance MHE / MPC scheme. While the control system is effective, its formulation is fairly complex, and holds too many control parameters. This makes its tuning and management probably more difficult than necessary. In this project, we will investigate different formulations of the MHE / MPC scheme, seeking simpler forms while retaining its control and optimization capabilities. The goal is to propose an MHE / MPC formulation that will then be applicable more broadly for smart house control.
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no)
Optimization-Based Guidance for Agile Satellite Attitude Maneuvers
With an increasing demand for readily available high-quality images, the maneuverability and pointing stability requirements of earth observation satellites continue to grow. SENER Aerospace is researching the application of convex optimization in attitude guidance and control to fully exploit the spacecraft capabilities and enhance their access rates to geographical locations of interest.
To contribute to this research, this thesis investigates optimization-based guidance strategies to be applied to the attitude control problem. Key challenges here include the nonlinear dynamics, but also pointing constraints to avoid the exposure of delicate instruments to harmful radiation. Of particular interest here is furthermore the use of a control moment gyroscopes as a torque device. While this can provide very high torque capacities, its control is typically challenging, e.g. due to internal singularities affecting the momentarily available torque.
The project will be conducted in collaboration with the company Sener Aeropspacial
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no)
To expand the capabilities of such systems, this thesis is to investigate optimization-based guidance strategies to be implemented onboard. A key challenge here are uncertainties and disturbances acting on the system – mainly in the form of winds – which should be taken into account in the generation of suitable trajectories. Since the resulting optimization problems can be computationally demanding, an implementation via suitable imitation learning techniques is to be investigated within this project.
The project will be conducted in collaboration with the company Sener Aeropspacial
NTNU ITK supervisor: Sebastien Gros (sebastien.gros@ntnu.no)
Scientific, academic and artistic work
A selection of recent journal publications, artistic productions, books, including book and report excerpts. See all publications in the database
Journal publications
- (2022) A Semi-Distributed Interior Point Algorithm for Optimal Coordination of Automated Vehicles at Intersections. IEEE Transactions on Control Systems Technology.
- (2022) Reinforcement learning-based NMPC for tracking control of ASVs: Theory and experiments. Control Engineering Practice. vol. 120.
- (2022) A new dissipativity condition for asymptotic stability of discounted economic MPC. Automatica. vol. 141.
- (2021) Can AI Abuse Personal Information in an EV Fast-Charging Market?. IEEE transactions on intelligent transportation systems (Print).
- (2021) Verification of Dissipativity and Evaluation of Storage Function in Economic Nonlinear MPC using Q-Learning. IFAC-PapersOnLine. vol. 54 (6).
- (2021) Risk-Based Model Predictive Control for Autonomous Ship Emergency Management. IFAC-PapersOnLine. vol. 53 (2).
- (2021) Reinforcement Learning of the Prediction Horizon in Model Predictive Control. IFAC-PapersOnLine. vol. 54 (6).
- (2021) Optimization of the Model Predictive Control Update Interval Using Reinforcement Learning. IFAC-PapersOnLine. vol. 54 (14).
- (2021) Bilevel Optimization for Bunching Mitigation and Eco-Driving of Electric Bus Lines. IEEE transactions on intelligent transportation systems (Print).
- (2021) The value of airborne wind energy to the electricity system. Wind Energy. vol. 25 (2).
- (2021) Optimal Model-Based Trajectory Planning With Static Polygonal Constraints. IEEE Transactions on Control Systems Technology. vol. 30 (3).
- (2021) Safe reinforcement learning using robust MPC. IEEE Transactions on Automatic Control. vol. 66 (8).
- (2020) A Game Approach for Charging Station Placement Based on User Preferences and Crowdedness. IEEE transactions on intelligent transportation systems (Print). vol. 23 (4).
- (2020) Data-driven Economic NMPC using Reinforcement Learning. IEEE Transactions on Automatic Control. vol. 65 (2).
- (2020) Experimental validation of a semi-distributed sequential quadratic programming method for optimal coordination of automated vehicles at intersections. Optimal control applications & methods. vol. 41 (4).
- (2020) Optimisation-based coordination of connected, automated vehicles at intersections. Vehicle System Dynamics. vol. 58 (5).
- (2020) Computing the power profiles for an Airborne Wind Energy system based on large-scale wind data. Renewable Energy. vol. 162.
- (2020) Optimization-Based Automatic Docking and Berthing of ASVs Using Exteroceptive Sensors: Theory and Experiments. IEEE Access. vol. 8.
- (2020) Reinforcement Learning-Based Tracking Control of USVs in Varying Operational Conditions. Frontiers in Robotics and AI. vol. 7 (32).
- (2020) Combining system identification with reinforcement learning-based MPC. IFAC-PapersOnLine. vol. 53 (2).