In a previous blog post, we shared how connectivity–particularly connected vehicles–can help power tomorrow’s Smart City. In this post, we explore how advances afforded by mobile technologies have radically changed our coverage of traffic network data, and how this helps to illuminate previously invisible city streets.
To develop plans for organizing mass transit routes, transportation planners need data like how long it takes to drive from A to B, how much traffic moves on a given street, and where a crash is likely to occur. But professionals have been in the Dark Ages when it comes to this critical work, relying on traditional technologies such as CCTV cameras, on-road sensors, speed radars, and on-board units (OBUs) attached to cars. Limited in network coverage, cost- and labor-intensive to install and maintain, these data sources meant that most of the traffic network in cities was technically invisible. For example, in New Jersey, surveillance systems only cover approximately 5% of the highway network. From a research perspective, modelers were forced to make a lot of assumptions, thereby increasing the risk of false conclusions in assessing road safety.
But thanks to the massive adoption of smartphones and advances in mobile technologies, real-time coverage of traffic networks is now possible. Although 150 million vehicles are expected to be connected via Wi-Fi by 2020 (according to Gartner), current penetration for connected cars is only 12% of the US car market. Contrast this with smartphone penetration of nearly 80% in the US (and higher in urban metros). Not only are smartphones more ubiquitous, they are also generally active. This is in contrast to an OBU on a connected car, which may remain inactive even when it is capable of sending data.
The advances in mobile technologies enable any smartphone to act like a connected sensor, thereby dramatically enhancing the quantity and quality of driver behavior data in a number of ways:
- Additional data: Unlike OBUs, smartphones can measure phone use, a critical contributor to distracted driving, which is responsible for 25% of motor vehicle fatalities today.
- Improved privacy: Collected data is crowdsourced and anonymized, so privacy concerns (a challenge with OBU where data remains within each vehicle) are readily alleviated.
- Better models: Because data is gathered from a “random” sample, there is more variability from different population segments, so models grow more robust and reliable.
- Growing validity: Not only are mobile technologies increasingly used to supplement traditional technologies, their validity continues to be more established each day.
- Immunity from cyber hacking: OBUs are susceptible to carjackers who have the ability to wirelessly hack into functions that can lower speeds, fully kill the engine, abruptly engage the brakes, or disable them altogether. Smartphones are immune from this cyber security threat.
Once invisible, “sensor-less” parts of the city can now be “illuminated” through mobile sensors. Even better, wide coverage can be accomplished in a cost-effective manner (We’ll cover that in the next blog post).
The map below visualizes the millions of trips analyzed in just one month by Zendrive’s smartphone-based driver analytics platform. We see massive amounts of driver behavior data on risky behavior–analyzing over 2 billion miles per month–without requiring additional equipment. More data leads to more accurate models, continuously improving algorithms and a lot less guessing.
Total number of trips Zendrive analyzed during November-December 2016 grouped by metropolitan statistical areas
How Zendrive data helps deliver better coverage for city planners
From a city planning perspective, Zendrive data can help identify risky spots and measure changes in road safety before and after projects and policies are implemented. (Check out our Mission Street study here.) Let’s say a city would like to identify risky intersections, but the only available data source is historical crash data (1-10 crashes maximum for the last 5 years at a single intersection). Since crashes are rare events, this is not unusual, but this is far from an ideal data size to derive a predictive model that measures riskiness at that intersection.
Zendrive measures risky driver behaviors–such as hard brakes and rapid acceleration–that can be interpreted as “near collisions.” This is invaluable for modeling location-specific risk factors. From a data-engineering perspective, it also makes sense for model-building since ultimately the goal is to eliminate leading indicators for crashes. With billions of near-collision events captured, Zendrive data offers a richness and granularity to predictive models that introduces a new dimension for city planning.
For governments and transportation agencies, it is also important to achieve a growth in data size that is in line with economies of scale. Collecting traffic data from smartphones technically does not require significant additional costs for planning purposes. In our next post, we’ll take a look at how the final “C”–cost-effectiveness–helps to enable Smart Cities.