How to build a crowdsourced traffic safety network

This is the second of a three-part series on the Department of Transportation’s plan to use mobile technologies to improve highway safety. Read part one and part three. 

The Department of Transportation’s decision to pursue building a nationwide traffic safety network presents planners with a challenge common to big public sector IT projects: picking a solution in an era of disruptive technology change.

DOT’s plan, for example, calls for using roadside and some in-vehicle sensors to send data on traffic conditions to various collection points. Yet the range of fixed sensors can be spotty, in some cases limited to locations where funding exists for installing the devices, experts point out.

Instead, some transportation technology researchers advocate the use of smartphone networks to collect data on traffic and road conditions. Yet even these applications have downsides. For one thing, using mobile apps for collecting traffic data depends on users having their apps turned on, which is not always a sure thing.

Is there a middle ground – with fewer trade-offs – between a fixed sensor or a smartphone network in setting up a future traffic safety network?

Trading in traffic data

A number of companies – including Google, Inrix and Cellint – think so. The firms are developing models for "trading" information with traffic data consumers. In practice, consumers would launch an application on their smartphones that displays crowdsourced data about traffic flows in the area. In return, the consumer's own location information – including speed and direction – would be transmitted to the company for analysis and inclusion in the traffic data set.

Companies are also working out similar agreements with public- and private-sector operators of vehicle fleets. They make money by selling the collected data to regional traffic-management systems. It's a deal that works well for everyone, says Shawn Turner, head of the mobility division of Texas A&M's Transportation Institute.

With this approach, "transportation departments don't have to put out for infrastructure," he noted. "And we've relied more on these private companies for data because they have the nationwide coverage that isn't there with the Bluetooth reader systems."

Jim Bak, public relations manager at Inrix, a provider of traffic data, said the firm, “started with the idea that if you could turn vehicles into sensors, or devices that travel with people in those vehicles into sensors, you could get traffic [data] not just where you install physical road sensors, you get traffic [data] everywhere cars go.”

“Aside from licensing the existing sensor data, the way we started to build a crowdsourced traffic network was to approach the commercial fleet companies who all had global positioning systems in their vehicles and had the ability to do two-way connectivity."

The deal was simple. "We started to create license agreements with these fleet companies: You tell us location speed and heading of your fleet vehicles, and we will give you traffic information in return that you can use for routing and logistics planning," said Bak. “Basic crowdsourced traffic was born."

Inrix has made agreements with auto companies – including Audi, BMW, Ford and Toyota – that offer built-in computer connectivity in some models of their cars. The company has also offered free downloadable applications – Inrix Traffic – for mobile devices that deliver real-time traffic data on Google Maps.

The applications, which track location data from the device's GPS, offer routing, estimated time of arrival and warnings of events and hazards.

Limitations of app-sourced data

While Bak acknowledges the value of adding smartphone apps to the crowdsource mix, he cautions that applications that aren't integrated with other data sources have severe limitations. With smartphone applications, he said, "you're really only getting data from people who have your app on and are using it in the car."

"Traffic can change at a moment's notice due to the weather, accidents, whatever," he said. "You want to have real-time data, you want to have lots of it and you want to have it refreshed as quickly as possible," Bak said.

Inrix's data sets are composed of approximately 60 percent fleet data and 40 percent consumer data. Currently, Inrix collects three kinds of data: location, heading and speed. And according to Bak, tests have shown that the company's reports of traffic speeds are within 5 miles an hour of the actual speed 98 percent of the time.

"The DOT and municipalities will use our real-time data in daily traffic operations in their traffic op centers," Bak said. "They'll also use that real-time data to supplement their traveler information services. The travel time messages in signs over the freeway in Virginia and Massachusetts are the result of our data."

Tapping the cellular network

TrafficSense, a traffic data service offered by Israel-based Cellint, uses a different strategy. Since the number of smartphones on the road running a particular GPS-based traffic application is going to be relatively small, Cellint looked for a way to tap into the movement of signals from all smartphones within range of a cellular network.

TrafficSense provides road traffic information based on tracking all active anonymous mobile phones over the road in a particular region. The system connects to the main switching centers of the cellular network and passively retrieves and analyzes all the signaling data at the switch, pinpointing the location of each active mobile phone in real time. This data is then used to generate the traffic information, including speed and incident detection and travel time between junctions.

TrafficSense sends cars out to map the signaling data over the roadway systems and creates a reference location database from that information. The real-time signaling data is then matched with the database to accurately locate each mobile phone to the exact road on which it is traveling and measure its exact speed on each road section. 

Our cars are on the road, recording the sequence of messages from the mobile unit to the network, said CEO Ofer Avni. "Every half a second when the phone is active, there will be a message going up on the network. We use the sequence of messages to create a signature of the road."

When Cellint receives real-time information from the network, TrafficSense uses pattern-matching analysis to track the exact location of the phone on the road. With a broader scale of moving sensors, Avni said, TrafficSense detects 99 percent of traffic slowdowns within a few minutes.

Next: Next generation apps on the road to smarter and safer highways