We have taken Bluetooth vehicle detection data and analysed it in a new way. Instead of considering simply how long it takes to get from one detector to the next, we have split the vehicles based on what is their next-next destination. This gives valuable insights into travel time with increased accuracy of what the journey time is.

An example from the video. A to B is along Main North Road at Thorngate from Park Terrace to Nottage Terrace (Scotties Motel). If you consider all traffic the average time at the 5pm to 6pm hour on the first day of data, the average time is 155 seconds. But if we consider vehicles that head North along Main North Road their average time is only 90 seconds. Compare this with those who turn right on to Nottage Terrace is 255 seconds. So the average that is quoted is an average for two streams of traffic with quite different experiences. The would show nothing the matter, or maybe a bit slow along this section of road, instead it should show, traffic flowing without delay heading north, but long delays to turn right.

The technique we have is used is to take the raw CSV files and then a number of Perl scripts to manipulate the data into triplets of traffic streams, each being A to B with an ultimate destination of C1, C2, C3 etc. These are then sorted and analysed by the final program This can all be automated by running the shell script The final output is a CSV file that can be viewed to show the road segment and then the average time. Blocks of data to the right of that show the deviation from the average time for each destination.

Team name
State, Territory or Country
Event location
Datasets used
Dataset Name
Detection of Vehicle-based Bluetooth devices (sample only)
Dataset Name
Bluetooth Detection Sites
Project image
Video URL (YouTube/Vimeo)