Multi-agent control under coupled spatio-temporal objectives with performance and feasibility guarantees
Event details
| Date | 25.09.2026 |
| Hour | 11:00 › 12:00 |
| Speaker | Dr Maria Charitidiou, Division of Decision and Control Systems at KTH and the Institute for Systems Research at the University of Maryland, College Park, USA |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
Abstract:
Over the years multi-agent systems have been involved in various applications in air, ground, sea and space offering increased performance and robustness compared to their single-agent counterpart. Nevertheless, designing performance-optimal control laws for teams of autonomous agents remains challenging especially when considering their limited communication, sensing and actuation capabilities. In this talk, we will consider spatio-temporally constrained tasks expressed in a formal specification language called Signal Temporal Logic (STL). First a centralized model predictive controller (MPC) is presented ensuring the satisfaction of the STL tasks, when possible or their minimal violation in presence of conflicting constraints. The proposed MPC problem accounts for the limited actuation capabilities of the agents, is integer-free and can be shown to be recursively feasible under appropriately designed terminal constraints. Next, decentralized and distributed MPC schemes will be discussed that guarantee the satisfaction of coupled STL tasks without however imposing any communication requirements among agents involved in the same tasks. A quantitative notion of trust will also be introduced that determines the agents’ trustworthiness according to their performance. This notion will be used to guide agents’ decisions towards satisfying collaborative tasks involving less trustworthy agents. We will conclude the talk with a recently proposed distributed control law that ensures asymptotic task satisfaction for tasks involving agents not capable of exchanging information directly. The proposed control schemes are applied to robotic and vehicle coordination scenarios, and their efficacy is verified in simulation.
Biography:
Maria Charitidou received her B.Sc. degree in mathematics from Aristotle University of Thessaloniki, Greece in 2015, the M.Sc. degree in systems and control from Delft University of Technology, the Netherlands in 2019 and her PhD degree in electrical engineering from KTH Royal Institute of Technology, Sweden in 2024. After her Ph.D. she held postdoctoral positions in the Division of Decision and Control Systems at KTH and the Institute for Systems Research at the University of Maryland, College Park, USA. Her research interests include formal methods, multi-agent systems, model predictive control and nonlinear systems and control.
Over the years multi-agent systems have been involved in various applications in air, ground, sea and space offering increased performance and robustness compared to their single-agent counterpart. Nevertheless, designing performance-optimal control laws for teams of autonomous agents remains challenging especially when considering their limited communication, sensing and actuation capabilities. In this talk, we will consider spatio-temporally constrained tasks expressed in a formal specification language called Signal Temporal Logic (STL). First a centralized model predictive controller (MPC) is presented ensuring the satisfaction of the STL tasks, when possible or their minimal violation in presence of conflicting constraints. The proposed MPC problem accounts for the limited actuation capabilities of the agents, is integer-free and can be shown to be recursively feasible under appropriately designed terminal constraints. Next, decentralized and distributed MPC schemes will be discussed that guarantee the satisfaction of coupled STL tasks without however imposing any communication requirements among agents involved in the same tasks. A quantitative notion of trust will also be introduced that determines the agents’ trustworthiness according to their performance. This notion will be used to guide agents’ decisions towards satisfying collaborative tasks involving less trustworthy agents. We will conclude the talk with a recently proposed distributed control law that ensures asymptotic task satisfaction for tasks involving agents not capable of exchanging information directly. The proposed control schemes are applied to robotic and vehicle coordination scenarios, and their efficacy is verified in simulation.
Biography:
Maria Charitidou received her B.Sc. degree in mathematics from Aristotle University of Thessaloniki, Greece in 2015, the M.Sc. degree in systems and control from Delft University of Technology, the Netherlands in 2019 and her PhD degree in electrical engineering from KTH Royal Institute of Technology, Sweden in 2024. After her Ph.D. she held postdoctoral positions in the Division of Decision and Control Systems at KTH and the Institute for Systems Research at the University of Maryland, College Park, USA. Her research interests include formal methods, multi-agent systems, model predictive control and nonlinear systems and control.
Practical information
- General public
- Free