Making robots (and humans) cooperate

University of Groningen Professor of Networks and Robotics Ming Cao has just received an ERC Consolidator Grant of EUR two million. His aim is to design algorithms that will help autonomous robots cooperate. These same algorithms may also help guide the decision-making processes of humans.

Autonomous robots and networked systems are hot topics. Self-driving cars are just one example of how intelligent systems can be used to create an autonomous ‘agent’. The car decides how fast it goes, checks for danger and adapts to its environment while the driver sits back and enjoys the ride. But if cars are to share the same stretch of motorway, it would be wise if they could talk to each other, share information and respond.

‘That’s why I want to study the evolutionary dynamics of networked autonomous systems’, explains Ming Cao. Autonomous agents make their own decisions on how to respond to the environment. But if these agents are in a network, all these individual decisions affect each other. ‘These processes are described by Game Theory’, says Cao. ‘The objectives of the autonomous agents are pre-programmed, but the way they achieve their objectives is not pre-determined. The agents make their own decisions.’

Cao wants to use the mathematics of Game Theory and Control Theory to produce an algorithm that will allow networks of autonomous agents to reach an optimal solution for an entire network. Take self-driving cars: ‘Commuters face traffic jams in the morning if too many cars take the same route. If some of them took an alternative but slightly slower route, this would reduce congestion.’

The algorithms Cao wants to design in the ERC project should produce evolving autonomous networks that can reach such optimized outcomes. ‘In five years’ time, I would like to have created a theoretical framework to design these networks. People could then use this to build their own applications. We will test these algorithms with teams of autonomous robots in our lab.’

His algorithms can also be applied to humans, to influence rather than control their behaviour. ‘Humans are autonomous beings, after all, so the mathematics describing their decision-making processes has some common features with those for autonomous robots.’ The algorithm could be used to optimize policies that are meant to encourage consumers to choose the sustainable option, for instance – or to reduce rush-hour traffic.

Cao will appoint four PhD students, each on a four-year project, and has five years’ funding for a postdoc. This will allow him to expand his research group, which now comprises some seven PhD students and one postdoc.

See also Sucking up electrons with acid