Keynote Speakers

From Self-Organization to Control: Steering Robot Swarms

Abstract: Robot swarms promise scalable and robust solutions for applications such as environmental monitoring, search and rescue, and warehouse automation, but their lack of manageability limits practical deployment. Fully self-organized robot swarms, operating without any central coordinating entity, have been widely demonstrated. In such swarms, the decentralized architecture provides high redundancy, and collective behavior emerges indirectly from local interactions. These features yield well-known advantages — such as scalability, fault tolerance through redundancy, and the absence of single points of failure — but also introduce fundamental challenges, most notably the difficulty of managing and influencing the swarm’s behavior. In contrast, centralized systems are easier to design, control, and manage, but suffer from single points of failure and limited scalability.

In this talk, I will begin by summarizing key results from fully self-organized robot swarms, highlighting both their strengths and their inherent limitations. I will then introduce a novel swarm architecture that achieves self-organized hierarchy, combining the strengths of decentralized and centralized approaches. Using a heterogeneous swarm of ground robots and aerial vehicles, I will show how the system self-organizes a dynamic hierarchical control network through local asymmetric communication. I will present experimental results demonstrating that the swarm can autonomously split and merge independently controlled sub-swarms, replace faulty robots anywhere in the hierarchy, and adapt its collective behavior in real time. Importantly, the proposed architecture preserves the key benefits of strict self-organization, including scalability and the interchangeability of individual robots. This approach provides a practical path toward deploying large-scale autonomous swarms in real-world scenarios where both robustness and controllability are essential.

Bio: Marco Dorigo is Co-Director of IRIDIA, the artificial intelligence laboratory at the Université Libre de Bruxelles (ULB), and a leading figure in swarm intelligence and swarm robotics. He is best known as the creator of Ant Colony Optimization (ACO), a landmark algorithm inspired by the collective foraging behavior of ants. First introduced in his 1992 doctoral work at Politecnico di Milano, ACO has since become a foundational method in the field of bio-inspired computation, with wide-ranging applications in routing, scheduling, and network design. Beyond optimization, Professor Dorigo has made major contributions to the field of swarm robotics, helping to shape the development of autonomous systems based on decentralized coordination and collective behavior. His work bridges theory and practice, advancing both the design principles and real-world implementation of intelligent, adaptive multi-agent systems. He is the Founding Editor of Swarm Intelligence and serves on the editorial boards of several leading journals in computational intelligence and robotics. Professor Dorigo has received multiple honors in recognition of his contributions, including the IEEE Frank Rosenblatt Award, an ERC Advanced Grant, the Marie Curie Excellence Award, and fellowships from AAAI, EurAI, and IEEE.

Provably Robust Artificial Intelligence? A Formal Methods Perspective

Abstract: The widespread deployment of AI is raising concerns about its robustness to manipulations, whether through adversarial or poisoning attacks, or strategic behaviour. Formal methods, and particularly formal verification techniques, are well placed to provide the needed assurance and the community has stepped into the fold. Using illustrative examples, this lecture will overview recent progress regarding the development of verification methodologies to provide provable guarantees of robustness, and outline future challenges.

Bio: Marta Kwiatkowska is Professor of Computing Systems at the University of Oxford and a Fellow of Trinity College. She holds degrees from the Jagiellonian University (BSc/MSc), Oxford (MA), and Leicester (PhD), and was awarded an honorary doctorate from KTH Royal Institute of Technology in Stockholm in 2014. Professor Kwiatkowska is a global leader in formal methods and verification, having pioneered probabilistic and quantitative approaches in verification that are now central to ensuring the safety and reliability of computing systems. She is best known for leading the development of PRISM, a software tool for probabilistic model checking, which has been widely adopted in academia and industry. PRISM allows for the analysis and verification of systems that exhibit uncertainty, such as those used in communication protocols, power management, and security, enabling the identification and correction of real-world flaws in these systems. Her current research focuses on the safety and robustness of machine learning and AI systems. Professor Kwiatkowska has been the recipient of various awards and honors, including the Royal Society’s Milner Award and Lecture—the first awarded to a woman—and the BCS Lovelace Medal. She is a Fellow of the Royal Society, ACM, EATCS, and Academia Europaea, among others. Her research has been consistently supported by numerous institutions such as the ERC, EPSRC, DARPA, and Microsoft Research, including two prestigious ERC Advanced Grants.