Research projects
Current projects
Equitable mobility management as a distribution control problem
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2027–2029
- Team
- Saverio Bolognani, Florian Dörfler, Kenan Zhang (EPFL)
Mobility management (tolls, speed limits, and other interventions that shape how people travel) is usually designed to optimize efficiency, for example to minimize the total travel time. Average metrics of this kind are blind to how benefits and costs are distributed across travelers, so equity is typically assessed only after the fact.
In this project we formulate equitable mobility management as a distribution control problem: the decisions of the planner shape the day-to-day travel behavior of the population, which in turn induces a distribution of utilities across travelers, and equity-oriented objectives (social welfare functions such as max-min fairness) are defined directly on this distribution.
We develop feedback control algorithms that optimize these objectives online from finite samples of travel data, with guarantees on convergence and sample complexity, and that are robust to the biased samples typical of real-world mobility data. The methods are demonstrated on real transportation networks.
Ethical automated decision making: From optimisation to sequential decisions
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2027–2029
- Team
- Saverio Bolognani, John Lygeros
Automation increasingly assists or replaces human decision makers, which raises the question of how to embed human ethical values in decision-making algorithms. When ethical duties conflict, moral judgement is usually left to the human in charge, a fallback that is not available when decisions are automated.
Building on our work on fairness and social welfare functions in multi-agent control, this project explores the connection between ethics and automated decision making in both directions. We investigate how ethical frameworks (utilitarianism, deontology, contractualism, virtue ethics) can be expressed in the mathematical formulations used in optimization, game theory, and stochastic decision making, and, conversely, what the choices made when formulating and solving these problems (relaxing constraints, regularization, dual variables and sensitivities) reveal about the ethics of the resulting decision.
We then turn to sequential decisions, where feedback can act as an ethical safety valve as uncertainty unfolds, but where decisions that seem ethical today may box the decision maker into a moral dilemma tomorrow.
Unleashing Local Energy Communities: Unlocking untapped value through automation and fair sharing principles
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2027–2029
- Team
- Saverio Bolognani, Line Roald
Swiss legislation allows neighboring households to form local electricity communities (LEGs) that trade energy locally through the distribution grid, with discounted network charges. These mechanisms are meant to encourage investment in local renewables, batteries, and flexible loads, but so far they have not worked as intended.
This project assesses the untapped value of local energy communities and investigates how to unlock it through automation and fair sharing principles. Using representative scenarios based on real data from the Grid2050 site in Walenstadt, we quantify the value that coordinated, automated operation can create compared to individual operation.
We then design game-theoretic sharing rules that align the decisions of individual members with the collective optimum, keep members willing to participate, and allocate costs and benefits fairly, and we survey prosumers to learn how they perceive these rules.
Finally, we assess the impact of LEGs on distribution grid operators and on consumers outside the community, and whether network discounts reflect true savings.
OPFV2: Automated incentives for voltage support procurement
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2025–2028
- Team
- Saverio Bolognani, Zhisen Jiang
This project develops adaptive control and incentive mechanisms for voltage regulation in the Swiss power transmission grid. As increasing penetration of distributed energy resources replaces large conventional power plants, maintaining transmission-level voltage requires closer coordination between the transmission system operator (Swissgrid) and distribution grid operators providing voltage support services.
The project investigates how control actions and financial incentives can be co-designed by combining feedback optimization, game-theoretic intervention, and data-driven modeling of the strategic behavior of distribution grid operators.
The proposed adaptive algorithms learn the state and flexibility of distribution networks from operational data and continuously improve voltage regulation performance through feedback. The methods are validated in large-scale simulations and on real operational data streams, with the ultimate objective of being incorporated into Swissgrid’s real-time decision-making and permanently integrated into its operational voltage control framework.
Selected publications
Karma Economies
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2024–2027
- Team
- Saverio Bolognani, Andrea Censi, Kai Zhang
Dynamic resource allocation problems, that is, decision problems where a contended resource needs to be allocated to agents based on efficiency and fairness criteria, can be attacked via “karma” mechanisms. When rational agents participate in this mechanism, an extremely efficient and fair allocation of the resource emerges, even though the agents are self-interested and there is no truthfulness assumption.
In this project, we go beyond the use of karma as a local resource allocation mechanism, and we look into its deployment in contexts with high societal relevance, where the decisions that are automated have consequences on society at large (either because they affect individuals or because they affect the environment where people live and then, indirectly, individuals).
Multi-agent control with limited interpersonal comparability
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2024–2026
- Team
- Saverio Bolognani, Heinrich Nax (UZH), Ilia Shilov
Multi-agent control and resource allocation problems require trading off the costs incurred by the different participants. This implicitly assumes that the gains and losses of different users can be compared: for example, that a traffic management scheme can weigh a few drivers with extremely high urgency against many drivers with minimal urgency.
In practice, this comparability is often limited or absent, because individual preferences cannot be revealed truthfully (for example, when monetary auctions are not allowed), because fairness principles require normalizing the agents’ preferences, or simply because preferences are hard to measure on a common scale.
Social choice theory has studied this question in depth, with famous impossibility results and precise prescriptions on which social welfare functions are admissible under which comparability assumptions, but these insights have been largely overlooked in control, where costs are often simply added up. This project brings these concepts into the design of distributed control and optimization strategies, opening a new avenue to introduce formal notions of fairness and equity.
Selected publications
Links
MAESTRO – Multi-Agent Energy Systems in Transmission grid Real-time Operation
- Funding
- Swiss Federal Office of Energy
- Period
- 2024–2027
- Team
- Saverio Bolognani, Giuseppe Belgioioso (KTH), Dominic Liao-McPherson (UBC), Jonas Matt
- Website
- maestro.control.ee.ethz.ch
Pervasive electrification and the massive deployment of renewable generation are increasing congestion in the power transmission grid. Rather than relying only on costly grid expansion, this project aims at a more efficient use of existing resources.
Solar plants, batteries, and controllable loads can provide complex real-time services (voltage regulation, congestion control, reactive power compensation, and even curative N-1 actions), but they are hidden in distribution grids: the transmission system operator has no direct control path to them, and they are managed by independent distribution system operators.
This project investigates the co-design of automated mechanisms and economic incentives to procure these services from distribution system operators.
We model the coordination as a large-scale Stackelberg game, in which the transmission system operator designs incentives and the distribution system operators respond through automated, feedback-based operation of their grids. The resulting protocols come with guarantees of efficiency, fairness, and stability, and are tested on a high-fidelity simulator built on real data from the distribution grid of Walenstadt.
Selected publications
Past projects
Dynamic population games for efficient autonomous mobility
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2021–2023
- Team
- Saverio Bolognani, Ezzat Elokda, Andrea Censi, Florian Dörfler, Emilio Frazzoli
We developed dynamic population games, a framework that couples the Nash equilibrium of a large population of agents with the dynamic equilibrium of their states, and showed that it is computationally tractable. We used it to design karma economies: non-monetary schemes that allocate shared mobility resources, such as road capacity at a bottleneck, fairly and efficiently. The work continues in the Karma Economies project.
Selected publications
- Dynamic resource allocation with karma: An experimental study.
Journal of Economic Behavior & Organization, vol. 251, p. 107771, 2026
[ paper: Dynamic resource allocation with karma: An experimental study | preprint: Dynamic resource allocation with karma: An experimental study ]
- A vision for trustworthy, fair, and efficient socio-technical control using Karma economies.
Annual Reviews in Control, vol. 60, p. 101026, 2025
[ paper: A vision for trustworthy, fair, and efficient socio-technical control using Karma economies | preprint: A vision for trustworthy, fair, and efficient socio-technical control using Karma economies ]
- Dynamic population games: A tractable intersection of mean-field games and population games.
IEEE Control Systems Letters, vol. 8, pp. 1072–1077, 2024
[ paper: Dynamic population games: A tractable intersection of mean-field games and population games | preprint: Dynamic population games: A tractable intersection of mean-field games and population games ]
- A self-contained karma economy for the dynamic allocation of common resources.
Dynamic Games and Applications, vol. 14, p. 578–610, 2023
[ paper: A self-contained karma economy for the dynamic allocation of common resources | preprint: A self-contained karma economy for the dynamic allocation of common resources ]
Links
Urban driving games
- Funding
- Swiss National Science Foundation (via NCCR Automation)
- Period
- 2021–2023
- Team
- Saverio Bolognani, Alessandro Zanardi, Andrea Censi, Florian Dörfler, Emilio Frazzoli
We modeled the interaction of autonomous vehicles in urban traffic as dynamic games, so that a vehicle can plan less conservatively by anticipating that the others react rationally. We introduced games with prioritized objectives (safety before progress before comfort), showed how large games factorize into smaller independent ones, bounded the efficiency lost to selfish driving, and developed sampling-based solvers for games with stochastic dynamics.
Selected publications
- How bad is selfish driving? Bounding the inefficiency of equilibria in urban driving games.
IEEE Robotics and Automation Letters, vol. 8(4), pp. 2293–2300, 2023
[ paper: How bad is selfish driving? Bounding the inefficiency of equilibria in urban driving games | preprint: How bad is selfish driving? Bounding the inefficiency of equilibria in urban driving games ]
- Posetal games: Efficiency, existence, and refinement of equilibria in games with prioritized metrics.
IEEE Robotics and Automation Letters, vol. 7(2), pp. 1292–1299, 2022
[ paper: Posetal games: Efficiency, existence, and refinement of equilibria in games with prioritized metrics | preprint: Posetal games: Efficiency, existence, and refinement of equilibria in games with prioritized metrics ]
- Urban driving games with lexicographic preferences and socially efficient Nash equilibria.
IEEE Robotics and Automation Letters, vol. 6(3), pp. 4978–4985, 2021
[ paper: Urban driving games with lexicographic preferences and socially efficient Nash equilibria | preprint: Urban driving games with lexicographic preferences and socially efficient Nash equilibria ]
UNICORN – A unified control framework for real-time power system operation
- Funding
- Swiss Federal Office of Energy, GRIDS program
- Period
- 2019–2021
- Team
- Saverio Bolognani, Lukas Ortmann, Jean Maeght (RTE), Patrick Panciatici (RTE), Florian Dörfler
- Website
- unicorn.control.ee.ethz.ch
The project took feedback optimization from mathematical theory to the real-time control of power grids, together with RTE, the French transmission system operator. We derived stability and robustness guarantees, released public benchmarks, and showed that a single feedback controller can handle voltage regulation, line congestion and re-dispatch better than today’s grid codes and industrial practice. The proposed control strategy has since been deployed on a real power distribution feeder in Switzerland.
Selected publications
Links
TEAMVAR – Networked feedback control of distributed energy resources for real-time voltage regulation
- Funding
- EU H2020 Research Programme, via the ERIGrid project
- Period
- 2017, 2019
- Team
- Saverio Bolognani, Adrian Hauswirth, Lukas Ortmann, Alexander Prostejovsky (DTU), Kai Heussen (DTU)
In two experimental campaigns at the SYSLAB facility of the Technical University of Denmark, we tested whether the power converters of solar and wind generators can regulate the voltage of a distribution feeder in real time. Purely local control could barely mitigate overvoltages, and sometimes made them worse. Networked feedback control, in which converters exchange information, unlocked their full reactive power capability with almost no model information, and in 2019 it ran fully distributed, without a central unit.
Selected publications
