On August 10th I spoke at the National Rural ITS conference at the Snowbird Resort in Utah. My goal was to push the audience to make better use of their work zone ITS data. Due to the length of my presentation I have broken it into two parts. But it is still much longer than my regular posts so I apologize in advance. I hope you enjoy it.
Thank you all for this opportunity to speak with you today. For those of you that don’t know me, I have been in the work zone ITS business for nearly 20 years. I started in the industry with ADDCO where I first learned about networking portable equipment to improve work zone safety. In 2001 I started Road-Tech to focus on this new and growing field of work zone ITS. But 20 years! It’s difficult to believe, isn’t it?
The primary goal of work zone ITS has always been improved safety and system efficiency, as it should be. But by deploying these systems we are now generating a great deal of data. It is time we began making better use of that data: both today and in the future.
This morning we will talk about data: about different kinds of data, and what works and what does not. We will talk about how best to collect that data and ways it can be used to improve the safety and efficiency of our work zones. We will talk about a new and innovative way to contract this work and, best of all, how to pay for it. Finally, I will share the exciting results of a recently released study proving the effectiveness of these systems. I want you to leave this session fired up about making better use of these systems, and knowing how to get that done.
So, why collect data? There are at least three reasons: 1) The Federal Work Zone Safety & Mobility Rule says we must collect data on our work zones and use it to evaluate their performance. In fact, the HSIP final rule describing these requirements will be published any day now. [ADD DETAIL AFTER FINAL RULE?]
2) By looking carefully at the data from current systems, we can find ways to make better use of the systems we are already using. They are like any other tool. When you fully understand its capabilities, you will make better use of it. This will result in even more improvements in the safety & efficiency of our roadways, and will make better use of our limited resources.
3) By learning more about the data generated, we learn more about the systems themselves and which ones work best in varying applications. Each manufacturer has their particular strengths and weaknesses. The easiest way to see those is through an examination of the data. This will make you better at defining what you need and want from systems on future projects.
But before you do, three things must happen:
1) You must have the right kinds of data…information you can use. In most cases this comes in the form of a log. The log will show reported speeds at each sensor with date and time stamps. It will show when a message sign was changed and what data triggered it.
2) You must collect it and hold onto it. And you must systematically categorize it for later analysis. There are data base programs available. Microsoft Access or something similar works well. But you could start with a simple spreadsheet and use that until you find you have outgrown it.
3) You must get to know your data, what affects it and how. We will talk about this in more detail later, but I will give you a simple example. On a queue warning system deployment, Caltrans noticed the message signs changing to warn of slow traffic early in the morning while volumes were quite low. They looked at the data and saw that one sensor was reporting stopped traffic while speeds at the other sensors were at the posted speed. They checked the job and found that the contractor was paving in that area and their trucks and rollers were the slow traffic. They relocated the sensor and the problem was solved.
So let us begin with data. Which data should we use? What works best and is most available? I spoke at a peer-to-peer exchange a couple of years ago about this very subject. At that time I asked far more questions than I answered. I think they were a little disappointed. But today the answers to those questions are more apparent. The list you settle on will include obvious metrics like traffic speeds and volumes, serious crashes, and perhaps delay time or queue length. Those are all measurable and reportable and are good indicators of conditions in your work zones.
You might also consider less obvious measures such as speed variance when speeding is a problem or volume reductions through the work area if your goal for the project was to divert traffic onto alternate routes.
There are several things you should consider when making these decisions. Let’s go through them one at a time.
1) Real time data versus collected data. This is obvious but bears discussion. Real time data is used to monitor current work zone performance and to trigger messages and other responses to changing conditions. Collected data is used later to evaluate work zone design, staging, and to plan future work zones in similar situations. Ideally the data you collect will work for both. But the needs for each will vary so be sure to check that all of your data needs will be met.
2) Project level data versus system performance data. The real data wonks in most agencies are accustomed to using system performance data. They want to know how well traffic flows over a single route or at a regional level. System data might tell us there was a problem in our work zone, but it would not tell us where. We need much more detailed, granular data so we can know what is occurring in and near our work zones. This is important because those systems operations folks will probably be leading any data collection effort. They are great advocates for data collection, but they don’t understand work zones. You must help them understand why you need something different.
3) Probe Data versus Spot Data. This is related to the last topic of project data versus operations data. Many agencies are knee deep in probe data. It is collected anonymously from cell phones, toll tags and BlueTooth. Cell phone data, in particular, is relatively inexpensive and easy to get. It is collected and displayed over a road segment. Some segments are as little as a couple of miles long but most are much longer. You may get lucky and find a segment covers the same stretch of road as a work zone, but not very often. So you end up with two or more segments overlapping your work zone or a single segment that’s much longer than you need. Warnings of slow or stopped traffic are only generated by probe data once average speeds over that segment drop below the trigger point. In other words, if you are only using probe data you may not learn about the incident for several minutes. Furthermore, probe data only tells you there is a problem. It won’t tell you where the problem begins. Spot data, on the other hand, tells you the moment speeds slow at a sensor. And because you have multiple sensors, you will know where in your work zone the problem begins.
4) The importance of “raw” data. The advantage of work zone ITS systems is the automation of responses to changing conditions. In that way we update message signs, send text alerts, etc. to warn of incidents and thereby reduce their impacts. It all happens instantly, without the need for human involvement. That’s great, don’t change a thing. But take the time later to look at the underlying data. When you have an incident, go back and look at what happened, when it happened, and where it happened. Compare the data to the outputs to learn how the system responded. I’ve always said that deploying these systems is 90% science and 10% art. The art comes from looking at the data and making adjustments to get the best possible performance from the ITS system.
5) The Importance of Multiple Data Points. One spot sensor might be all you need if you are worried about traffic backing up on an off ramp, or for a trucks entering system. But most of the time you need multiple sensors. For queue warning we recommend spacing them a mile to as little as a half mile apart. The smaller the distance between them, the faster you will learn of any problems. That means you will respond faster resulting in shorter delays, and fewer secondary crashes. For travel or delay time systems, more data points result in more accurate reporting.