MetroLab Network has partnered with Government Technology to bring its readers a segment called the MetroLab Innovation of the Month Series, which highlights impactful tech, data and innovation projects underway between cities and universities. If you’d like to learn more or contact the project leads, please contact MetroLab at firstname.lastname@example.org for more information.
In our first 2019 installment of the Innovation of the Month series, we explore how the University of Central Florida, in Orlando, is using big data and analytics to predict and mitigate traffic accidents.
MetroLab’s Executive Director Ben Levine spoke with Mohamed Abdel-Aty, Pegasus Professor and Chair of the Department of Civil, Environmental and Construction Engineering at the University of Central Florida (UCF); Dean Michael Georgiopoulos, College of Engineering and Computer Science at UCF; Charles Ramdatt, director of Special Projects in the city of Orlando; and Jeremy Dilmore, Florida Department of Transportation engineer, to learn more.
Ben Levine: Dr. Abdel-Aty, you’re using data to mitigate traffic accidents. Can you tell me about your project?
Mohamed Abdel-Aty: Certainly. My team at the University of Central Florida has been utilizing data sources for real-time crash prediction for many years. This effort initially started in partnership with the Florida Department of Transportation and the Colorado Department of Transportation. The most recent research for this work grew out of the U.S. Department of Transportation’s Solving for Safety Visualization Challenge, for which our work has been selected as a Stage-I semifinalist. As part of that challenge, we stated that by integrating real-time and static data, we could develop predictive analytics to diagnose real-time traffic safety conditions.
The improvement of analytics software and emergence of a data-rich environment contribute to the data-driven analysis for traffic safety by investigating crash, traffic, weather, geometric data, etc. Traditionally, traffic safety data analyses were conducted based on static and highly aggregated data, like annual average daily traffic or annual crash frequency. These aggregated data analyses can only reveal the general trend and relationship between crash frequency and few contributing factors, which could result in unreliable findings simply because they are averages and cannot represent the real conditions at the time of a crash.
The input data are the foundation to conduct real-time data-driven analysis for road safety. In recent years, with the advancement of big data, abundant data could be used for better crash prediction. My team has utilized these data — including Microwave Vehicle Detection, Automatic Vehicle Identification, Bluetooth, and real-time weather — on both arterials and freeways, including at typical conflict areas (e.g., intersections, weaving areas, ramps, etc.), to develop accurate algorithms that can predict the increase in crash risk in real time.
Bringing all these tools and algorithms under one integrated system will enable operators to monitor safety risk in real-time and develop interventions that alleviate the potential problems and prevent crashes or at least mitigate their severity.
If readers would like to learn more, they are welcome to visit my Google Scholar homepage for more than 200 publications addressing these concepts and proctive traffic management.
Levine: I suppose it’s obvious to readers that we’d like to reduce crashes. But perhaps you could offer some more detail about the root causes of the challenges you are addressing and how government can respond.
Abdel-Aty: The advent of big data technologies enables real-time analysis for traffic safety. By integrating multiple data sources, the data could help us understand the relationship between the presence of traffic conflicts and real-time contributing factors (e.g., volume, speed, traffic control status and weather characteristics), and quantify the impact of these factors on real-time crash risk. It is well-known that most crashes happen because of the presence of conflicts between different road users. For example, the conflicts between a leading and a following vehicle could lead to rear-end crashes. To date, different real-time countermeasures have been deployed to avoid conflicts.
The availability of microscopic detailed data is a major enabler and the possibility for the first time to prevent or reduce severity of crashes in real time.
Jeremy Dilmore: Visualization tools that make safety information accessible and understandable allow us to operationalize the insights. This gives us the ability to go from being reactive to proactive to determine the need for countermeasures prior to any unsafe trends developing along roadways. This represents a major step in our ability to deliver a fatality-free roadway system. The department is doing this in partnership with the city of Orlando and the University of Central Florida. The team believes this is the way of the future.
Levine: What kind of data are you exploring and why? What have been some of your initial findings in your research, and is this changing how you view the issue?
Abdel-Aty: Figure 1 shows the data that should be considered as input, including crash data, traffic data, traffic control data, road quality and design data, and other data like weather condition and land use. Traffic data from various detectors could provide all the traffic variables such as volume, average speed, average lane occupancy, and standard deviation of these parameters during specific time intervals to represent the traffic conditions and turbulence prior to crash occurrence. Traffic control data could provide real-time traffic control statuses and help to predict traffic conditions in the future time intervals. Other environmental information could be obtained from the roadway characteristics inventory database, a local climatological data set, etc.
Figure 1. Data for Crash Prediction (available and currently used by Dr. Abdel-Aty’s team). Courtesy of Dr. Abdel-Aty.
Our extensive research has proven the validity of the methodologies that we have developed for five-minute increments (predicting the safety risk for the next five to 10 and 10 to 15 minutes). We can select appropriate intelligent transportation systems techniques, including adjustment to adaptive signals on arterials, route diversion to/from freeways or arterials, variable speed limit, and ramp metering on freeways before crash/conflict risk increase, and even the advanced connected vehicle technologies. Last but not least, extensive simulations have proven the concept of real-time evaluation and visualization for various real-time risk mitigation techniques.
Levine: How has your work impacted city planning?
Abdel-Aty: Our more recent work focuses on arterials, city streets and intersections, which will have a major impact on the improved traffic flow and safety on city streets.
For a specified road (city street, arterial or freeway), a spatial-temporal screening technology (city streets in Figure 2) could be used to present the crash risk along the road by time and segment. For the selected diagnosed locations, we could use both of the screening methods to present the crash risk prediction results in real time for the selected roadway sites.
Figure 2. Rear-End Crash Risk for different locations and time along a street segment. (Dimitrou, Stylianou and Abdel-Aty, 2018). Courtesy of Dr. Abdel-Aty.
Charles Ramdatt: UCF has an excellent, experienced, team of academics, that is rightly focused on this important topic. I look forward to the benefits that will accrue to metropolitan Orlando, which is often ranked among the least safe metropolitan areas for pedestrians. Also, the metro area, with a population of almost 2.75 million, has over 72 million annual visitors. This means that during critical weeks of the year, the visitor population causes the population of the area to virtually double. UCF’s work in this atypical “traffic laboratory” is, therefore, of critical importance.
Levine: How has this project shaped your view of emerging tools and data analytics methods?
Abdel-Aty: First, I now better appreciate the power of the data and more advanced analytics methods. We can accurately predict the increase in crash risk with well-trained machine learning approaches. I believe we need to develop tools to show decision-makers the potential, so they are more favorable to adopt these prediction and pro-active traffic management methods.
Michael Georgiopoulos: Dr. Abdel-Aty and his group have been pioneers in adaptive traffic management approaches using real-time data and big data analytics. Their work has shown the potential to improve traffic safety in real time, which is a big deal. It is no surprise that their work has been chosen by the U.S. Department of Transportation as one of the semi-finalists in Solving for Safety: Visualization Challenge; UCF being the only institution in the semi-finalists of this challenge is a testament of the innovative work of Dr. Abdel-Aty and his team.
Levine: Where will this project go from here?
Abdel-Aty: Active traffic management is the most prevalent method for real-time road safety management, which primarily includes ramp metering, variable speed limit, dynamic lane management, integrated corridor management, etc. However, the traditional active traffic management strategies are usually triggered by a specific incident or event, which may still result in traffic delay. In this context, my team proposed proactive traffic management based on real-time crash risk prediction results, which are able to proactively prevent the potential incident or crash occurrence. Among our previous studies, ramp metering and variable speed limit have been widely investigated for real-time crash risk reduction on freeways, including at weaving areas (Figure 3). For example, Abdel-Aty et al. (2007) evaluated the expected benefits of using a modified ramp-metering algorithm as a method for real-time safety improvement on an urban freeway. We found that there are significant benefits in metering multiple ramps when the feedback ramp metering algorithm is implemented at multiple locations and increasing the number of metered onramps produces increasing safety benefits. Moreover, Yu and Abdel-Aty (2014) proposed an innovative approach to identify an optimal variable speed limit control strategy with the purpose of proactively improving traffic safety on freeways. Results indicated that a variable speed limit would effectively improve traffic safety under high and moderate compliance levels.
We are now bringing all the real-time safety prediction algorithms and proactive traffic management strategies together in an integrated comprehensive system that will enable operators to manage and improve traffic safety and operation in real time. We are also moving from active to proactive traffic management.
Figure 3. Safety analysis for weaving area (Wang and Abdel-Aty, 2016, 2017). Courtesy of Dr. Abdel-Aty.
Link to original post