As roadside hardware is now capable of more advanced processing, the need to send large amounts of data to the cloud for analysis is reduced. But how useful can this be? We asked Dr Subramanian Ramamoorthy, associate professor at the School of Informatics, University of Edinburgh.
In regard to traffic signals, what are the benefits of moving from a system where data is processed in the central server or cloud, to one where it is processed locally?
The main benefit of allowing for more local decision making is the possibility of responding to local circumstances, for example, unusual congestion that may not be captured in an initial pre-programmed sequence.
To what extent will the autonomous vehicle revolution be a catalyst for more edge processing?
Autonomous vehicles provide unique opportunities for improving the infrastructure by going digital. In a connected ecosystem where cars and the infrastructure could talk to each other, you can imagine obtaining efficiencies that were simply impossible if all cars were driven by humans, who are only able to respond to a coarse level of signaling, and slowly at that.
Self-driving vehicles will need to process as well as generate lots of floating data to function. To what extent is this the reason why traffic lights will need to be able to process data locally?
Consider what currently happens with Google Maps and Waze in terms of map data, and particularly traffic data – they use real-time feeds from many different phones to actively provide a nowcast of what is going on within the road network. Such data could equally well be obtained by the traffic management systems. If so, it enables the provision of richer and more useful representations of the state of the road – especially at difficult intersections and other hard to navigate regions where neither traffic rules in the abstract or ones’ immediate sensors are sufficient to obtain full situational awareness. By the time we have self-driving cars on the road, these could be valuable in enabling safety, which then makes it necessary that we allow for such data processing and sharing.
With the IoT and the physical internet set to dominate over the next two decades, processing data through ‘edge’ systems will become more prevalent. Will it spell the end of the central processing, or will central systems continue to be utilized?
I think that the edge and the central nodes have different roles. The edge will focus on the ‘now’, whereas a central store could focus on extracting longer-term knowledge and making that available to the edge – for example, features or models that are useful to the edge, extracted from big data aggregated across numerous roads and cities, which a single edge node will neither have access to nor possess the resources to work with.
How long will it take for the majority of traffic installations to process data locally? And how will traffic management centers that receive the data need to change?
This I do not have enough insight into. There is a desire in the context of connected and autonomous vehicles to make this happen. For some approaches to AVs, it may even be the best way to achieve safe driving. However, there are many more socio-political factors that would determine ultimate adoption of technical ideas.
For more on the ‘revolution at the edge’ don’t miss the April/May 2018 edition of Traffic Technology International.
By James Gordon