Presentation - Learning Interactions in Multi-Robot Systems

Systems composed of multiple robots can perform tasks that are beyond the capabilities of a single robot. Multi-robot and multi-agent systems are widely established research fields, and are experiencing broad uptake across industry sectors. Yet how are we to orchestrate teams of agents? How do we distill global goals into local robot policies? Machine learning has revolutionized the way in which we address these questions by enabling us to automatically synthesize decentralized agent policies from global objectives. In this presentation, I describe how we leverage data-driven approaches to learn interaction strategies that lead to coordinated and cooperative behaviors.  I will introduce our work on Graph Neural Networks, and show how we use such architectures to learn multi-agent policies through differentiable communications channels. I will present some of our results on cooperative perception, coordinated path planning, and close-proximity quadrotor flight. To conclude, I discuss the impact of policy heterogeneity on agent alignment and sim-to-real transfer.

Biography

Amanda Prorok is Professor of Collective Intelligence and Robotics in the Department of Computer Science and Technology at the University of Cambridge, UK, and a Fellow of Pembroke College. Prior to joining Cambridge, Amanda was a postdoctoral researcher at the General Robotics, Automation, Sensing and Perception (GRASP) Laboratory at the University of Pennsylvania, USA. She completed her PhD at EPFL, Switzerland. 

She has been honored by numerous research awards, including an ERC Starting Grant, an Amazon Research Award, the EPSRC New Investigator Award, the Isaac Newton Trust Early Career Award, and several Best Paper awards. Her PhD thesis was awarded the Asea Brown Boveri (ABB) prize for the best thesis at EPFL in Computer Science. She serves as Associate Editor for Autonomous Robots (AURO).