'Machine learning' demonstrated at innovITS Advance

Research carried out by the University of Southampton into the application of ‘machine learning’ approaches to the optimization of traffic signal control, has been screened on prime-time UK television. The university’s research has been using computer games and simulations to investigate the main attributes of good traffic control. The work has shown that, given the right conditions, humans are excellent at controlling the traffic, but that it is difficult to get a computer program to perform at the same level. To demonstrate its finding, the traffic lights at one of the main junctions of the UK’s ITS research center, innovITS Advance, circuit were controlled first by a computer and then by a human operator, while 30 volunteer drivers tried to negotiate the junction. The result was a win for human-based control of the signals, with longer delays for computer control. The university’s researchers have now developed ‘machine learning’ traffic control computers that can ‘learn’ how to control the lights in the same way as a human being and can improve their own strategies through experience.

Dr Simon Box of the University of Southampton Transport Research Group explained, "In transport research we are always looking ahead, and we can consider a future where all vehicles are equipped with WiFi and GPS and can transmit their positions to signalized junctions. This opens the way to the use of artificial intelligence approaches to traffic control, such as machine learning. The demonstration carried out at innovITS Advance for BBC’s ‘The One Show’, indicates that the human brain, carefully employed, can be an extremely effective traffic control computer. In our research we aim to be able to emulate this approach in a new kind of software that can provide significant benefits in improving the efficiency of traffic flow, hence improving road space utilization, reducing journey times and potentially, improving fuel efficiency.”

Operations manager at innovITS Advance, Catherine Ferris, commented, “We were pleased to be able to host this demonstration by the team from Southampton University. The development of artificial intelligence-based approaches to junction control is one of many new and promising technologies that can help us make better use of existing urban and road capacity, while reducing the environmental impacts of road traffic. The highly controllable innovITS Advance circuit, where roads, WiFi mesh, mobile phone networks, and even GPS reception, can be controlled in a manner that would be impossible on the public roads, provides an excellent environment in which such new technologies can be developed and perfected, before their commercial implementation.”

28 August 2012