Graph Neural Networks for Decentralized Multi-Robot Path Planning by Qingbiao Li, Fernando Gama, Alejandro Ribeiro and Amanda Prorok won Publication of the Year in the Hall of Fame Awards 2021.
The Hall of Fame is the group of companies set up by students, staff and alumni of the Department of Computer Science and Technology – and there are now over 300 of them. “We are very proud of this,” says Ben Karniely, coordinator of our alumni association, the Cambridge Ring. “It is an indication of the impact that our Department has on the computing industry and the world of technology.”
The 2021 awards ceremony took place on Thursday 7 April at a Cambridge Ring dinner at Queens’ College Cambridge. The awards are given in four categories:
- the company of the year;
- the product of the year;
- the publication of the year (from among those produced by members of the Department);
- and the Better Future Award – which is intended to recognise those who have made significant contributions to humanity through technology.
The paper Graph Neural Networks for Decentralized Multi-Robot Path Planning addresses the issue of learning to coordinate autonomous agents through graph neural networks (GNNs). The problem of coordinating autonomous agents is notoriously hard but Professor Amanda Prorok and her team, with colleagues at the University of Pennsylvania, tackled it by proposing a learning-based solution that leverages a GNN to produce decentralized decision-making policies.
They demonstrated the power of their solution on the problem of multi-agent path finding: the method performs near-optimally, at a fraction of the computational cost when compared to state-of-the-art benchmarks. They also showed that the method scales to arbitrarily large problem instances. The video that accompanies the academic paper has over 1,500 views on YouTube.
Comments are closed.