Autonomous cars have a potentially fatal flaw: They struggle to detect and react to cyclists on the road. According to a January 2017 report by IEEE Spectrum, bicycles are generally considered “the most difficult detection problem that autonomous vehicle systems face.” It’s not surprising: Bikes are relatively small, nimble, and sometimes unpredictable, and human drivers have a hard time sharing the road with bike riders as well. In 2015, 818 cyclists died in collisions with motorists, and 45,000 experienced injuries in car-bike collisions. In 2016, the number of deaths rose to 835. Seventy-one percent of those happened in urban areas.
It’s a multifaceted problem. Some drivers haven’t been educated as to how to properly share the road with cyclists, some cyclists don’t know how they’re supposed to behave, and infrastructure in many places doesn’t facilitate peaceful coexistence on the same roadways. Autonomous vehicles, with their advanced sensing capabilities and predictable, programmed behavior, offer the opportunity to help change that. However, we’re increasingly learning that A.I. can amplify our own biases and human failings. If humans aren’t doing a good job of detecting and preventing vehicle-bike collisions, how can we create machines that do the job even better?
One solution presented by Ford, Tome Software, and Trek Bicycle at CES last month is a concept known as bicycle-to-vehicle communications. Instead of just autonomous vehicles (or all motorized vehicles) on the road being able to wirelessly communicate their position and intentions with one another, bikes would be able to join the party. The proposed technology would be brand agnostic, something any cyclist could affix to herself or her bike. The key safety aspect of this connectivity would be that drivers would be alerted when a cyclist is nearby. It’s similar, although potentially a step above, a concept presented by Volvo in 2014 that would work through tech embedded in a rider’s helmet. Tome plans to hone its software, which could then be licensed out to vehicles, apps, bike accessories, and car accessories, at the Mcity autonomous driving test facility at the University of Michigan over the next year.
The concept is great. You can picture an autonomous car, the reach of its sensors expanding out into the world in concentric waves, detecting the ping of a cyclist’s signal at its back right corner and moving left to give it three feet of leeway in the lane. It could also save valuable seconds of processing time on the autonomous vehicle’s part. “Instead of spending time analyzing a series of images looking for people on bicycles, with a less than perfect rate of success, a car can simply make requests for devices in its vicinity and within milliseconds understand its environment,” an article in mountain-biking publication Singletrack explains.
There’s one problem: This is cheating. Autonomous cars, out there beta testing on U.S. roads today, can accurately detect other vehicles, pedestrians, even big game charging suddenly across a street. Forcing cyclists alone to strap a sensor onto their backs feels like a crutch, a cop-out. In truth, some envision that the only way an autonomous car future will work is if everyone—vehicles, cyclists, pedestrians, pets—are connected to the same system, an idea known as V2X, or vehicle-to-everything. This would help self-driving vehicles to be the most informed about their surroundings and prepare for any possible interactions. The problem is that it requires everyone to take part, which poses several noteworthy financial and logistical questions, such as who pays for this system, how it’s deployed, how it’s enforced, and whether pedestrian and traffic laws would need to be changed in order to facilitate cooperative behavior. (For example, stricter jaywalking laws to ensure pedestrians only cross in places self-driving cars expect them to.) In this scenario, autonomous car success hinges on a large number of difficult-to-control variables. But if the cars themselves are able to successfully sense and react to their surroundings, from a cyclist taking the lane to a toddler dashing into the street, the only variable that needs controlling is the technology itself.
Luckily, while some self-driving car companies may be looking at bicycle-to-vehicle communications as an option, many in the advanced stages of testing don’t appear to be cutting corners. Take Waymo for example: The Alphabet-owned company has been working on cyclist-detection technology for several years, and its cars have reportedly avoided collisions in several tricky scenarios, including giving a rider extra room when a parallel-parked car’s door opens and stopping suddenly when a rider rounded a corner and turned straight into oncoming traffic. Uber, Lyft, and even Apple have expanding fleets on our roadways, and they wouldn’t be out there if company executives didn’t believe these vehicles could avoid vehicle-bike collisions—at least most of the time.
Bicycle-to-vehicle communication is a good idea and could be useful in certain scenarios, such as when visibility is low—at night or in the rain—or on tricky, twisty back roads with blind corners. But if cars are going to drive the roads without human help, they need to be able to handle all of the challenges that come with it, regardless of whether they’re wirelessly connected to the world around them.
Support our independent journalism
Readers like you make our work possible. Help us continue to provide the reporting, commentary, and criticism you won’t find anywhere else.