The staggering cost of inaccurate data – Traffic Technology International

March 1, 2020

Traffic Technology International | March/April 2020

The staggering cost of inaccurate data

In the 1990’s, Dr. Pravin Varaiya, a distinguished professor at the University of California at Berkeley, was working on some exciting developments in the field of traffic analytics when he realized that the prerequisite for everything envisioned by the industry was a level of data accuracy very rare in the existing systems deployed globally. This was the impetus for the development of a wireless detection platform that has leveraged recent progress in ultra-low-power sensing and wireless chips to deliver the required level of accuracy and reliability.

Evaluating detection options

The traffic industry today has a broad choice of competing technologies for detection and data collection including inductive loops, wireless sensors, video, radar, tubes and probe vehicle data, each with its advantages and limitations; as no single technology can claim to be the best answer for all applications. However, there is general agreement that for real- time data applications, in- pavement solutions like wireless sensors and inductive loops are more accurate, while video and radar are less intrusive but
significantly less accurate.

When choosing a technology, practitioners are faced with the following question: How accurate does a solution have to be to meet the requirements of my application, and how much of a cost premium should that accuracy command? In other words, what is the cost of inaccurate data? As we show in the following example, that cost can be surprisingly high if the data is used for timing traffic signals.

Need to know
The benefits of wireless sensors over inductive loop technology

> Wireless sensors save installation time and maintenance costs over inductive loops.
> Minimal lane closures are required and is safer for workers on the road
> Roadways are not damaged by sensors and detection is not destroyed by resurfacing.

Suppose a traffic signal controller runs on fixed time plans where signal splits (i.e. division of cycle time into durations for each phase) are computed from vehicle counts or flow estimates. Suppose also that the estimated flow in one phase is in error by E, that is, the actual flow is (1+E)xF, where F is the estimated flow in vehicles per hour. If the intersection is functioning at capacity (i.e. cycle time just accommodates the total traffic), then on average the affected phase will result in ExFxT/3600 vehicles being queued for each saturated cycle, where T is the cycle time in seconds. These queued vehicles each will wait an additional T seconds. The result is an extra delay of ExFxT2/3600 vehicle- seconds per saturated cycle. If P percent of the cycles are saturated then the daily extra delay is: (ExFxT2/3600) x(24*3600/T)xP vehicle-seconds per day = (ExFxTx24xP)/3600 vehicle-hours per day.
If, for example, the actual flow is undercounted by 10% (E = 0.1), T= 100 sec, the nominal flow F is 3,000 vehicles/hour, and P is 25%, then the extra delay is 50 vehicle-hours per day. At $20 per vehicle-hour this is equivalent to $1,000 per day. So the decision to choose a technology that is, say, $5K per intersection cheaper but 10% less accurate ends up costing the community a staggering $365K per year in lost productivity.

This simple analysis assumes a fixed time signal plan. If the signal is adaptive, the cost of inaccuracy will be even larger. Similar analysis for other applications like actuated signals, ramp metering, active lane management or dynamic pricing applications will readily demonstrate that the cost of inaccurate data can be extremely high. The cost of inaccuracy is not lost on experienced traffic agencies and is the key reason why inductive loops remain by far the most widely used technology for vehicle detection globally for nearly 30 years. These loops are in the roadway and provide lane-specific data without all of the concerns of occlusion from objects, lighting conditions, and weather.

Wireless sensors are a fast growing alternative to inductive loops because they offer a host of advantages while retaining the accuracy level loops established. Sensors require a fraction of the installation time resulting in much less time in the lanes, less danger for workers, and much less disruption for the public.

Additionally, only the lane where the sensor is being installed needs to be closed, unlike loops where multiple lanes may need to be closed simultaneously to run the wire to the side of the roadway.

Left: In-ground detection is widely used for its accuracy
Right: Accurate vehicle monitoring keeps costs down

Wireless simplicity

Wireless systems make it easier to reach locations without power and/or communications such as mid-blocks on arterials and freeway lanes and ramps where power may not be available. In addition, wireless systems benefit from comprehensive remote monitoring, diagnostics, and alerts for all system components which significantly cuts the maintenance costs and enables preventative work which reduces public complaints due to detection not functioning.

And finally, sensors operate on a single comprehensive software platform for numerous applications with common hardware components, thus achieving operational and cost efficiencies. Ultimately, using a wire-based loop in our roads today is similar in concept to still using a corded phone.

The cost of inaccurate data in terms of lost productivity can be staggering. Our simple example shows that a 10% count inaccuracy at a highly traveled
intersection can result in more than $365K per year in lost productivity. Only in-pavement technologies like loops or wireless sensors deliver the accuracy needed for advanced traffic management solutions, but with their significant advantages such as safety, ease of installation and maintainability, wireless sensors are rapidly become the preferred solution globally.