Skip to Main Content

Bicyclist fatalities in 2015 were at their highest level since 1995–818 cyclists died and approximately 45,000 were injured.

To reverse this alarming trend, Anthony Rowe, an associate professor in electrical and computer engineering at Carnegie Mellon University, says that early-warning collision systems that are making their way into our auto fleet must not only detect cyclists but also predict how they will move.

Rowe, in collaboration with Bruno Sinopoli, associate professor of Electrical and Computer Engineering, conducted research to create a figurative “protective bubble” or proximity around bikes with respect to human-driven cars and future self-driving cars. Proximity, or the safe space between cars and bikes, can be driven by legislation. In Pennsylvania, for example, cars have to stay four feet away from bikes. Yet, rules are not enough to protect cyclists. Accidents occur because humans often don’t pay attention when driving. Even when drivers are alert, it can be difficult to see cyclists, like at dusk. Biking during the evening rush hour is precarious, too. Rowe tailored his work to examine how smart technologies could intervene and prevent collisions between bicycles and cars.

Rowe and Sinopoli, both avid cyclists, led a team that designed a hyper-instrumented test bike. The bike was outfitted with sensing devices, such as cameras, GPS, accelerometers, and dedicated short-range communication (DSRC) radios that sent data to a smartphone that was attached to the bike. Rowe gathered a variety of information including the speed of the bike, the proximity to obstacles around it, the distance between it and cars, the motion of the bike, etc.

Using the data they collected, they can estimate how far into the future a computer could predict the trajectory of the bike with respect to surrounding cars on the road. Predicting how a bike will move was not easy. Cars generally follow a fixed route down the street. Cyclists, on the other hand, may ride on the edge of the road alongside traffic or ride in the middle of the lane like a car. Cyclists may even jump up on a sidewalk mimicking a pedestrian.

After modeling how the bike’s trajectory could be predicted, Rowe says the next steps entail the cyclist’s smartphone sending signals nearby cars. To illustrate, after a bike’s sensors assess the surrounding traffic activity, the cyclist’s smartphone will send a DSRC signal to an approaching car’s collision-warning system indicating that an accident will occur in two seconds if the car doesn’t slow down.

It is anticipated that in the not-too-distant future DSRC capabilities will be installed in all new cars because the technology will improve safety by alerting drivers in real time about road hazards. In addition to vehicle-to-vehicle (V2) communications, or as Rowe’s work highlights bike-to-car communications, DSRC will be used for vehicle-to-infrastructure (v2i) communications. These intelligent transportation systems technologies have the capability to improve the safety of everyone—the drivers, cyclists and pedestrians—who use our roads.