In the world of traffic and transportation engineering assumptions are abundant. Some informed and some not so much. Luckily, guidelines and standards backed by rigorous research have for years helped make these assumptions a lot more informed. While not ideal, data scarcity has been the main contributor to this practice. An example of these assumptions is capacity related variables in Level of Service (LOS) analysis. Any engineer or analyst in this area has found themselves making assumptions regarding saturation flow rates for signalized intersections and the variables that impact it. These assumptions however, are not harmless. In fact, a study conducted by researchers at Purdue University¹ quantified such impacts in terms of their effect on LOS analysis with surprising results.
Through a monte carlo simulation approach a standard deviation of 180 vehicles was considered varying around the true saturation flow rate. The results of their study is summarized in the figure provided below. The curve represents the distribution of LOS classification compared to the true LOS value. As an example in a true degree of saturation of 0.95, with the error in saturation flow rate a true LOS of D is only predicted correctly in 42% of cases!
The effect of these assumptions are obvious but the question is, is there a better way? With advances in connectivity and turn key solutions in unlocking data from sensors available in the field, real measurement of performance has never been this straightforward. Automated performance data can help validate or reject assumptions, effectively allowing data to show us how different methods and strategies perform. For example, below is a congestion scan visualization of travel time along a corridor with 4 major intersections showing travel times compared to free flow throughout the day. The intersections are configured to operate in free mode throughout the entire day. The data clearly shows decent operation throughout the entire day as a trend over several weeks except for a very specific couple of hours on the NB direction during the PM peak.
From the controller’s perspective, despite free mode operation, cycle lengths are turning out to be quite high. In fact the intersections are operating at +120 second very frequently.
From a coordination perspective during those hours things aren’t looking good either. Green bands are non-existent and downstream green starvation is obvious. The data is clearly showing that the free mode operation is not benefiting anyone.
The point is to use the data to understand if the free mode operation objectives are being achieved or not under real world circumstances by answering questions using data. Questions such as: In which hours throughout the day the current timing plan is achieving our objectives and which hours aren’t? Is a slight reduction in minor movement delay in receiving green time worth the high delays along the corridor? What are pedestrian delays and would it significantly increase if intersections are coordinated considering that cycle lengths are already quite consistently high? What hours would coordination be most beneficial? What cycle length should be used for coordination? And many more.
The data can help answer all these questions with minimal need for assumptions. The Miovision TrafficLink software puts all of these analytics at you fingertips without worrying about the need for any custom spreadsheets, databases or data engineering and science capabilities. Being data driven isn’t hard, you just need the right tools.
- Tarko, Andrzej P., and Marian Tracz. “Uncertainty in saturation flow predictions.” Red 1 (2000): P2.