Self-driving cars have a number of ways to “see” the world around them—not only with cameras but also with lidars, radars, and other sensors. Many self-driving cars also have microphones that give them a sense of hearing, allowing them to detect honking horns, blaring sirens, screeching tires, and so forth.
An Israeli startup called Mobi-Wize is aiming to give cars—self-driving and otherwise—a sense of touch.
Rather than installing a new suite of sensors in vehicles, the company aims to draw on sensors already in vehicles—most notably wheel-speed sensors, along with data on the position of the gas and brake pedals, steering wheel, and so on—to draw conclusions about road conditions.
Mobi-Wize CEO Amit Nisenbaum described his technology as “a virtual sensor implemented in software… It’s an application of AI and machine learning on data that is coming from the existing sensors,” he explained.
One of the most important data feeds provided by the Mobi-Wize software is how well the vehicle’s tires are gripping the road. “Mobi-Wize’s technology reliably estimates, in real time, the quality of the contact between the vehicle and the road,” the company’s website says. That gives the car a specific, numerical estimate of how quickly it can accelerate or decelerate without worrying about skidding.
Conceptually, this is similar to what anti-lock brakes and traction control systems do on today’s cars. But Nisenbaum told me that Mobi-Wize’s more sophisticated software-based approach has some big advantages. Conventional anti-lock brakes kick in after a car’s wheels have already started to slip, he said. By contrast, Mobi-Wize aims to predict available traction ahead of time, so the car can apply the maximum braking force that won’t cause the wheels to lock up.
Nisenbaum says the company is working with carmakers that are evaluating whether its software can improve the performance of conventional traction-control and anti-lock braking systems. We will have to wait for independent analysis to see whether that idea pans out.
The real breakthrough: Crowdsourcing road-quality maps
The potential applications for the technology go far beyond improving conventional traction control and anti-lock braking systems, however. That’s because Mobi-Wize software collects data about road conditions and uploads it to a cloud-based service for later use.
For several years, self-driving car companies have been building high-resolution maps recording the precise location of curbs, lane dividers, signs, and other objects in and near the road. Mobi-Wize is doing something similar, but it focuses on the physical characteristics of the road surface.
Rather than sending out fleets of mapping cars, Mobi-Wize plans to crowdsource this data from customers running its software. Nisenbaum said the company has already amassed more than 10 million kilometers (6 million miles) of road data. The company has the most complete coverage in its native Israel, but it also has significant data in parts of Germany and expects to build maps for other parts of the world over time.
There are a lot of potential applications for a map like this—especially for self-driving cars. If a pothole is up ahead, the car can shift its lane position slightly to avoid it. If rain or snow starts to fall, the Mobi-Wize software should be able to predict which parts of the road will ice over the fastest. Cars will be able to adjust their speed and following distance based on the precise characteristics of the road ahead.
In the heart of the city
Another possible application Mobi-Wize is working on: “Municipalities can use these capabilities to do planned maintenance,” Nisenbaum told us.
Mobi-Wize already has pilot projects underway with municipalities, he said. The company has supplied cities with aftermarket devices that hook into the ODB ports in their existing vehicles. As city officials go about their regular business driving around the city, Mobi-Wize collects data about road conditions in the background.
Then Mobi-Wize can provide city officials with detailed maps of every road that is suffering from potholes or other signs of deterioration. This detailed and objective data could enable cities to focus maintenance efforts on locations that need it the most.
Cars in the future will use many different types of sensors to understand the world around them—this is just the latest example. Last month we covered a company trying to bring military-style ground-penetrating radar to market to help cars deal with snow. The month before that, we covered a startup creating a “virtual sensor” to help autonomous vehicles understand the intentions of pedestrians.
Early self-driving cars won’t have the kind of sophisticated understanding of the world that human drivers take for granted. That insight allows humans to drive a car using little more than a pair of cameras we call our eyes. By contrast, self-driving cars will likely have access to a multitude of different kinds of data about the environment—including a number of sensor types not available to human drivers.
