Communication-constrained Cooperation under Uncertainty: How to Make Strategic Decisions?

Overview

In many real-world systems, multiple agents must cooperate while facing uncertainty about the environment, each other’s capabilities, and the consequences of their actions. Communication between agents is often limited by bandwidth, latency, privacy, or cost.

Led by Setareh Maghsudi at the University of Tübingen, the Cyber Valley Research Fund project “Communication-constrained Cooperation under Uncertainty: How to Make Strategic Decisions?” (2021–2024) investigated how agents can make effective strategic decisions when they must cooperate under such constraints. The project combined ideas from game theory, decision theory, and machine learning to model and analyse distributed decision-making in networked multi-agent systems.

Key results

• Explored negotiation strategies and federated-learning scenarios where participation is limited by communication costs or financial constraints, highlighting trade-offs between model accuracy, communication overhead, and the number of participating agents.

• Developed game-theoretic and learning-based models for cooperation in multi-agent systems with incomplete or uncertain information.

• Studied how limited or delayed communication affects the ability of agents to sustain cooperative behaviour, showing that communication constraints can shrink the “punishment horizon” and make cooperation harder to maintain.

• Analysed distributed task allocation and resource-sharing problems in cyber-physical systems and drone-based networks, modelling them as cooperative games under state and incentive uncertainty.

• Investigated how emergent communication policies learned by reinforcement-learning agents influence cooperation in social-dilemma settings, finding that communication can enhance cooperation under symmetric uncertainty but may also enable deception under asymmetric uncertainty.

Why it matters

The project provides theoretical and algorithmic foundations for designing multi-agent systems that can cooperate reliably despite uncertainty and limited communication. This is relevant for applications such as autonomous drone fleets, smart grids, distributed sensing, and federated learning, where agents must coordinate without full information and under communication constraints.

The results help clarify when and how communication supports cooperation, when it can be exploited, and how system designers can structure incentives and communication protocols to promote robust collective behaviour. As a primarily theoretical and algorithmic research project, it does not deliver a single deployed system, but rather principles and models that can guide the design of future cooperative AI and robotic systems.

This project produced the following peer-reviewed publications: 

• Mittal, V., Maghsudi, S., & Hossain, E. (2021). Distributed Cooperation Under Uncertainty in Drone-Based Wireless Networks: A Bayesian Coalitional Game. IEEE Transactions on Mobile Computing, 20(12), 3458–3472. doi:10.1109/TMC.2020.3001856

• Cao, Y., Maghsudi, S., & Ohtsuki, T. (2021). Mobility-Aware Routing and Caching: A Federated Learning Assisted Approach. In IEEE International Conference on Communications (ICC) 2021. doi:10.1109/ICC42927.2021.9500678

• Maghsudi, S., & Lan, A. S., Xu, J., & van der Schaar, M. (2021). Personalized Education in the Artificial Intelligence Era: What to Expect Next. IEEE Signal Processing Magazine, 38(3), 37–50. doi:10.1109/MSP.2021.3055032

• Ghosh, B., Haghighi, M. H., Motagh, M., & Maghsudi, S. (2021). Using Generative Adversarial Networks for Extraction of INSAR Signals from Large-Scale SENTINEL-1 Interferograms by Improving Tropospheric Noise Correction. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, V-3-2021, 173–180. doi:10.5194/isprs-annals-V-3-2021-173-2021

• Maghsudi, S., & Davy, M. (2020). Computational Models of Human Decision-Making with Application to the Internet of Everything. arXiv:2008.11958 [cs.MA].

• Ehyaei, A.-R., Karimi, A.-H., Schölkopf, B., & Maghsudi, S. (2023). Robustness Implies Fairness in Causal Algorithmic Recourse. In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’23), 1233–1250. doi:10.1145/3593013.3594007

• Balef, A. R., & Maghsudi, S. (2023). Piecewise-Stationary Multi-Objective Multi-Armed Bandit with Application to Joint Communications and Sensing. arXiv:2302.05257 [cs.LG].

• Khosravi, M., & Maghsudi, S. (2026). Robust Optimization Approach and Learning Based Hide-and-Seek Game for Resilient Network Design. arXiv:2602.11854 [cs.NI].

• Davoodi, M., & Maghsudi, S. (2025). Efficient Resource Allocation under Adversary Attacks: A Decomposition-Based Approach. arXiv:2506.23442 [cs.GT].

• Maghsudi, S., & collaborators. (2024). Emergent Cooperation under Uncertain Incentive Alignment. arXiv:2401.12646 [cs.AI].

• Dai, S., Maghsudi, S., & Thiele, L. (2024). Gradual Change Detection in Covariance Matrix: A Lazy Approach. In IEEE Wireless Communications and Networking Conference (WCNC) 2024.