NCCR Automation collaborative project
Dynamic population games for efficient autonomous mobility
The goal of this collaborative project is to demonstrate that multiple competitive agents can efficiently share a mobility infrastructure without the need for an external coordinator (in other words, autonomously). We plan to do that by developing a new theoretical foundation and new computational methods to study and design population games with dynamics. In this new formalism, the concept of Nash equilibrium is coupled with the concept of dynamic equilibrium. Our preliminary investigation shows how this approach can produce automated and scalable solutions for the coordination of mobility agents, while improving fairness and efficiency in the use of the available resources.
NCCR Automation collaborative project
Urban driving games
One of the hard problems that prevent the deployment of Autonomous Vehicles (AVs) is the problem of motion planning in an uncertain dynamic populated environment, where the AV must take into account the (uncertain) presence of other vehicles and their (uncertain) intent. Only recently we understood the necessity of studying this problem also from the perspective of game theory. The naive solution that is the state of the art (planning assuming the others behave according to certain models and considering a worst-case analysis) is extremely conservative; an AV should instead be allowed to plan more aggressively, knowing that the other AVs will react rationally. In this project we study dynamic (multi-stage) decision-making problems that represent the interaction of multiple cars in an urban environment, we look for structural features of these games, and we develop computational tools to solve them.
A unified control framework for real-time power system operation
The goal of this project is to engineer a unified control solution for power systems that replaces many real-time control mechanisms and emergency procedures (secondary/tertiary frequency control, economic re-dispatch, line congestion control, voltage control), enabling the participation of active distribution networks as flexible providers of ancillary services. The proposed approach is based on a novel mathematical method for the design of feedback control laws that steer the power system state towards safe and efficient working points, ensure satisfaction of the grid operational constraints during the resulting transient, and guarantee the closed-loop stability of the grid dynamics. The resulting real-time strategy can outperform today's best practices and to ultimately enhance the system capacity to host intermittent renewable energy sources.
Networked feedback control of distributed energy resources for real-time voltage regulation
Future power systems will be characterized by a large penetration of renewable energy sources, typically characterized by intermittend and partially unpredictable behavior, and a distributed charging infrastructure for plug-in electric vehicles. It is believed that current power distribution networks will need a structural reinforcement in order to host these new classes of consumers while ensuring that the complex physical constraints of the grid (voltage limits, power line capacity, voltage stability) are satisfied. This project challenges such idea. Its goal is the development of control strategies for the real-time actuation of the grid based on the measurements obtained from a distributed sensing infrastructure, exploiting the unused flexibility of the available power converters. The experiments proposed in this project aim at testing two claims: 1. Communication between converters is necessary for effective voltage regulation 2. Scalable distributed communication architectures are as good as centralized ones. The experimental validation is performed on a proof-of-concept prototype at the Technical University of Denmark.