Shaping Emergent Behaviour Through Integrated Physical AI

Abstract

Robots and autonomous systems are increasingly deployed in society, from automated vehicles on urban roads to drone fleets in low-altitude airspace, forming large collective systems of interacting agents. In such settings, intelligence that resides solely in a centralised “brain”, abstracted away from the robot body and its dynamics, rarely delivers the intended system behaviour at scale in the physical world.

In this talk, I will argue that resilient real-world autonomy lies at the intersection of three complementary forms of intelligence rather than in any one of them: computational intelligence, which learns and decides; embodied intelligence, in which the body and its physical dynamics share the computational load; and collective intelligence, which emerges from interactions among many agents. A unifying lens for all three intelligences is physical reservoir computing, a framework that treats the natural dynamics of physical systems as a computational resource. Drawing on examples from large-scale mobility systems, I will show how the physical dynamics of swarm systems can act as computational resources and how they can be leveraged to shape emergent behaviour toward desired outcomes while preserving resilience, adaptability, and scalability. I will conclude the talk by discussing future research directions for aeronautical and aviation engineering.

Speaker