Abstract: Correct traffic light phase sequences are critical to efficient traffic flow through road
intersections. If the timing of just one set of lights at a junction is one second from
optimal, the results can lead to major traffic congestion and delays. Currently, traffic
light phases are set using a mixture of experience and empirical techniques to achieve
an optimal phase sequence and, in the worst cases, this process can take many
months. Any technique to speed up the identification of optimal phasings would have
clear benefits to government agencies everywhere.
Thales UK’s Research and Technology Group have investigated how Genetic
Algorithms can solve “real-world” problems and have obtained some promising results
with optimising traffic light phasings. Working with their colleagues at Thales Research
and Technology Netherlands they have developed a demonstration solution based on
the concept of “software agent communities” – that is, groups of software programs
working together for their greater good. This extends the earlier Genetic Algorithms
work by using agent-based technologies to enable the rapid optimisation of larger
traffic light networks, potentially covering an entire town.