University of Cambridge > Talks.cam > Microsoft Research Cambridge, public talks > Real time control with lots of humans in the loop

Real time control with lots of humans in the loop

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Microsoft Research Cambridge Talks Admins.

This event may be recorded and made available internally or externally via http://research.microsoft.com. Microsoft will own the copyright of any recordings made. If you do not wish to have your image/voice recorded please consider this before attending

Did you know that humans can control traffic light junctions better than the leading automatic adaptive control systems? The optimal switching policy for traffic lights on a network of junctions is computationally intractable (EXP-time complete) and current systems use approximate optimisations to determine the switching policy. Embodied simulation experiments and also computer based simulation experiments, which employ a “traffic control computer game” have shown that human controllers can significantly outperform these automated controllers on many metrics. Furthermore by applying supervised learning to the traffic control computer game it is possible to capture player’s strategies and thus new trained automated controllers with high performance have been developed.

A key difficulty in traffic control is that any signal control system is not the only control system ‘in the loop’. Each vehicle is controlled by an independent human who is not aware of (or even motivated towards) the actions they could take for optimal system performance. Furthermore, assuming that a benevolent controller had this information, their available control actions are only weakly acting (changing the traffic lights at junctions). In fact there are several other control problems that share these properties: Crowd control, queueing problems, building evacuation, battlefield operations. In these “lots-of-humans-in-the-loop” systems real-time control actions are limited or absent and optimal control strategies are computationally intractable.

In this talk I will discuss the potential for building on the traffic control computer game to develop large multi-player games for simulating lots-of-humans-in-the-loop systems and ‘gaming’ scenarios where real time control is applied. As in the traffic control computer game the developed control strategies may be in part learned from human players thus exploiting the powerful intuition and empathy of humans in solving the apparently intractable problem of how to control them.

This talk is part of the Microsoft Research Cambridge, public talks series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2017 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity