Data-driven intersections help ease traffic flow: Las Vegas traffic signal optimization
RTC’s next-generation transportation network
Geotab along with Waycare and the Regional Transportation Commission of Southern Nevada (RTC), are collaborating to create a first-of-its-kind product that utilizes data-driven analytics and actions to support traffic signal operations.
Each partner contributes a unique set of skills to the collaboration:
- Geotab anonymously aggregates data from over 60,000 connected vehicles in the region that generate more than 2 billion records per day; Geotab mashes this data up with a myriad of third-party sources and applies the necessary machine learning and analytics to generate traffic insights that characterize traffic signal performance.
- Waycare integrates the Geotab data directly into its AI-based traffic management platform and creates predictive models that inform RTC traffic signal operators.
- Staff at the RTC traffic management center use the results to create traffic signal timing plans that measure performance related to safety, efficiency, and throughput for travelers and trip purposes.
For over a year, we have gathered a huge amount of data from connected vehicles. Now we are in the process of combining the capabilities of Waycare while using Geotab’s data-driven insights to determine how to solve problems based on the accumulation of mass data. By implementing automation into our strategy that has been proven through advanced modeling, we will see net benefits to the flow and safety of Las Vegas traffic.Brian C. Hoeft, P.E., Director of Traffic Management Center for RTC of Southern Nevada
Data-Driven traffic signal coordination schemes
Traditional traffic engineering efforts rely on traffic simulation tools to evaluate signal infrastructure performance. However, the simulated results are only as good as the data and corresponding analytics that are supplied. As an example, traffic flow diagrams such as the one shown in Figure 1 below are used by traffic signal operators to coordinate traffic along a targeted corridor for the purpose of balancing safety with efficiency and throughput. This is achieved by implementing progressive traffic signal phases that allow vehicles to travel as platoons through a corridor without stopping.
Figure 1: Simulated traffic flow diagram
In Figure 1, the platoon speed (indicated in blue) is a critical input to traffic coordination design as it dictates the required cycle lengths, offsets and splits for the corridor. Various data collection techniques are used to calculate median free flow speeds as a proxy for platoon speed, however these traditional methods typically do not capture the statistical distribution in vehicle speeds along the targeted corridor. Similarly, emerging techniques such as the Purdue Method for Signal Performance, which focuses on discrete data obtained from the traffic signal software and hardware, are also ineffective at exposing this distribution. As a result, Geotab and RTC are developing a new approach to traffic signal coordination that combines signal system data with connected vehicle insights to greatly enhance signal timing efficiency.
As an example, see below for a real distribution of vehicle speeds along the Cheyenne Ave corridor in Las Vegas, running between North Rancho Drive and the Las Vegas Freeway (Interstate 15). This bimodal distribution is a typical one as vehicles naturally cluster into larger slower moving vehicles or smaller faster moving ones. This information is a critical input into any traffic signal progression model. The next step is to identify the peak periods for heavy-duty and light-duty vehicles and coordinate signals accordingly. This data may inform us that it is more valuable to favor light-duty vehicles during the traditional morning and afternoon commute times versus heavy-duty vehicles during the very early mornings and midday.
Figure 2: Cheyenne Avenue corridor
Figure 3: Histogram of average vehicle moving speed
So what platoon speed should we use? And when should it be used? The Cheyenne corridor is a great example of a corridor that has a higher proportion of heavy-duty vehicles, compared with other corridors that emphasize personal or consumer vehicle trips. By coordinating signals for a statistical average or median speed (28 mph for example), traffic management staff are not directly prioritizing either vehicle class. Similarly, by prioritizing light-duty vehicles, heavy-duty vehicles are indirectly punished. See Figure 4 for an illustration of this conundrum.
Figure 4: Traffic flow diagram showcasing bimodal speed distribution
This trend is further illustrated in RTC’s trajectory plots shown in Figure 5, where you see actual vehicle trajectories and travel profiles through the Cheyenne Ave corridor. Note that some vehicles take much longer than would otherwise be expected to traverse the corridor.
Figure 5: RTC trajectory plots and travel profiles through Cheyenne corridor
The simple analysis above illustrates the complexities associated with coordinating traffic signals along a straight corridor. This exercise becomes exponentially more complicated when analyzing an entire traffic network with temporal variations in congestion and speeds. To optimize signal timings and plan infrastructure changes across the entire network, a multivariate analysis is required that relies on rich transportation insights. As demonstrated by Geotab, connected vehicles are an amazing source of data necessary to produce these insights.
Intersection and corridor insights: Cheyenne Avenue case study
Geotab helps produce insights from aggregate data across millions of commercial and consumer connected vehicles to characterize traffic flow and corridor performance across signalized intersections. These insights help uncover:
- Average dwell or stop times per intersection per time of day per turning movement
- Traffic queue lengths per intersection per time of day per turning movement
- Number of signal cycles per queue (how many green lights to clear a queue)
- Corridor travel time
- Corridor stop propensity
In the era of big data and the internet-of-things, with an increasingly large number of connected vehicles we are able to generate insights that were not possible before. Leveraging Geotab products with cloud infrastructure and machine learning techniques, we can create smarter cities that provide more data and enable us to continually improve.Daniel J. Lewis, Senior Data Scientist at Geotab
These insights can be used to benchmark turn signal efficiency versus straight-through signal efficiency, specifically when running historical analysis to find unique daily or hourly trends. Using this data, RTC is able to assess the effectiveness of certain signal phase changes within a targeted corridor, and more importantly across their entire transportation network. See below for a detailed analysis of a specific intersection along the aforementioned Cheyenne Avenue corridor: Cheyenne Avenue and Losee Road.
Figure 6: Number of stops to clear intersection
For optimal throughput there should be one or less stops on average to effectively service vehicles through the corridor. Figure 6 shows that for most directions of travel through this intersection the goal of one or less stops on average is being met. However, for South to East left turn from Losee Road onto Cheyenne Avenue, for a substantial portion of the day it is taking more than one stop for a vehicle to clear the intersection. This could indicate that the capacity for the left turn lane is being exceeded. It also presents some other questions that can now be more easily addressed:
- Is it good for 11 of the 12 movements to have average or good performance and one to have poor performance?
- Should the agency construct additional capacity for this turning movement?
- If the same analysis was performed at several intersections, could the agency prioritize where the most needed improvements were?
- Once capacity was changed, how was the performance impacted?
Figure 7: Average time from first stop to clear intersection
Figure 7 shows the average time from first stop to clear the intersection is much higher for left turns heading South West on Losee. It also shows extended wait times turning left off of Cheyenne heading East and onto Losee heading North East at certain times of the day. The through directions are likely prioritized for this corridor as we see much lower time from first stop for East and Westbound traffic on Cheyenne, with slightly more congestion in the Eastbound direction.
This allows us to objectively evaluate questions or complaints related to traffic that sits through the longer waits. Is it worth it to coordinate signals to provide so much time to through movements at the expense of other travelers? Given the direct communication between the traffic center and the Las Vegas Valley signal network, a signal timing adjustment that shifted time from the through traffic to the turn traffic could be implemented.
Figure 8: Average distance to first stop
Figure 8 shows that for a large part of the day, the average distance to first stop approaches 200m for several movement patterns through the intersection. This is the case for Losee Road SW turning left onto Cheyenne heading E. The lane length for this left-hand turn is only 200m long indicating capacity for the lane is likely exceeded, and in turn the left-hand queue runs the risk of backing into straight-through lanes and impacting flow in the prioritized straight-through direction. The signal timing could potentially be adjusted to provide slightly more time for vehicles turning left, while taking into consideration the queue length in other directions which would be affected.