Selective Vehicle Detection (SVD) ? Bus Priority and GPS Technology

Selective Vehicle Detection (SVD) ? Bus Priority and GPS Technology


N B Hounsell, S Ishtiaq, F N McLeod, University of Southampton, UK; K Gardner, S Palmer, T Bowen, TfL Surface Transport, UK



Selective Vehicle Detection (SVD) to achieve active bus priority at traffic signals has achieved impressive results particularly in terms of Urban Traffic Control techniques known as SPRINT and Bus SCOOT. Comprehensive SVD route treatment can reduce end to end bus journey times by over 2 mins with junction delays reduced by 22% - 33%.

Transport for London (TfL) is investigating the use of GPS technology in public transport for bus location and related services such as selective vehicle detection at Traffic Signals. Although GPS-based systems are perceived to be flexible and versatile, a drawback of this technology is the locational error associated with it. This error could result in some buses missing the given priority actions, particularly extensions. This paper provides analysis and discussion of the effect of GPS error on bus priority taking account of different detector locations, detector combinations and operational conditions.

This study was carried out using TRG's microscopic simulation model, SIMBOL. Theoretical analysis was also carried out to increase the robustness of the results. Simulations were carried out with the assumptions that buses are detected using virtual detectors which may be positioned anywhere on the link. GPS location errors were assumed to be random and unbiased and GPS detection was assumed to be available 100% of the time. Bus priority was assumed to be awarded as done in SCOOT.

Simulations of single virtual detector at various locations for different levels of GPS error showed that:
Increasing GPS error reduces the bus priority benefits. In general, the reduction in benefit due to GPS error was found to quite small when compared with the impacts of other factors such as the junction degree of saturation, using central extensions and using a lower value of the SCOOT parameter bauth (the maximum displacement of a stage change allowed for an extension).
The reduction in bus delay savings due to GPS error appeared to be fairly constant over the different detector distances considered. A possible explanation for this is that there was a trade-off between the effects of having more priority extensions at the longer detector distances, which would tend to increase the impact of GPS error, and having greater bus journey time variations at the longer detector distances, which would tend to dampen or mask the impacts of GPS error.
Existing bus journey time variability influences the impact of GPS error on bus delay savings. Simulation results showed that the influence of the GPS error is more noticeable where the journey time variability is low and less noticeable where the variability is high.

Results from using an exit detector to end priority extensions when the bus is detected showed that the GPS error influences the optimal location of the exit detector. Assuming the use of local extensions, the GPS error required the location of the exit detector to be shifted downstream of the signals by a distance equal to the maximum detection error anticipated. This was necessary to avoid premature termination of extensions when buses are detected before they pass through the signals.

Results from using an exit detector to hold priority extensions until the bus reaches the signals, in addition to the cancelling function, increased bus delay savings slightly. The increase in bus delay savings came from the reduction in the number of buses missing extensions. Further results were obtained for runs where the bauth value was increased. An increased bauth value did not produce any benefits here, however, as the existing bauth value was sufficient to cover the average bus journey time and busvary.

The benefits of using a secondary detector were found to be limited in this research, where bus stops were not considered. The use of a secondary detector will be of most benefit at higher levels of bus journey time variability between the primary and secondary detectors. One of the most useful situations for using a secondary detector occurs where there is a bus stop close to the traffic signals and the primary detector is placed upstream of the bus stop. Additional simulations supported this by showing higher bus delay savings with multiple detectors for higher bus journey variability on the link. If GPS virtual detection can enable secondary detection at little or no added cost, then it could be worthwhile implementing secondary detectors on all links where the primary detector is more than 100m, say, from the signals.

Results were obtained for three different levels of junction degree of saturation: 60%, 80% and 95%. These showed similar trends with bus delay savings tending to reduce with increasing degree of saturation. The impact of the junction degree of saturation was shown to be greater than the impact of GPS error. The impacts of using central extensions and of using a lower value of bauth were both shown to be greater than the impacts of GPS error on bus delay savings.

Overall, the simulation results showed that GPS error, of the magnitudes expected, has a relatively low impact on bus delay savings when compared to the impacts of other factors such as the junction degree of saturation, use of central extensions and use of a lower value of bauth. Earlier research work in UG94 (Bretherton et al, 1999) also showed that errors in location positioning of up to 10 metres have little effect on bus journey time savings. The commonly reported GPS accuracy of a maximum error of 5m appears to be sufficient for bus priority purposes if combined with a bus door closing dead reckoning..


Association for European Transport