Everyone's been there: driving in the pitch dark, attempting to decipher signs, handle sharp turns, and weave through multiple lanes of whizzing traffic. It's a difficult situation even for an experienced human driver--so how can a car that's driving itself pull it off?

An autonomous vehicle must know its precise location, the location of other cars around it, the route to its destination, and any possible obstacles in its path. To deliver that information, automakers are turning to technology developed by San Francisco-based startup Civil Maps. The 30-employee company says it can give cars data that's more accurate and more frequently updated than competing self-driving systems.

Civil Maps' software guides a vehicle using cameras, lidar (light detection and ranging) sensors, and computer vision. This cognition system detects the shape, size, and movement of nearby objects such as stop signs, signals, and pedestrians. It then uses the massive amount of data to produce detailed maps of street-level infrastructure, directing the car to make safe decisions. The software's interface also gives passengers a sense of control by showing them the information the car is receiving from the environment on an augmented-reality map.

 inline image

Self-driving cars aren't what Civil Maps co-founder and CEO Sravan Puttagunta set out to pioneer. Just three years ago, the Indian-born, Stanford- and Berkeley-educated Puttagunta was working on digital fingerprinting--technology that allows music and other media rights holders to monitor copyright infringements--at San Francisco-based television startup Samba TV. "I realized fingerprinting technology is very good at pattern recognition," says Puttagunta. "And the underlying information theory practices could be applied to autonomous vehicles."

In the three years since its inception, Civil Maps has secured $6.6 million in funding from investors including Ford, which took part in its seed funding round last July. Other backers include Motus Ventures, a fund focused on connected mobility, and Stanford's StartX accelerator. Puttagunta says the company has annual revenue in the "low seven figures." It expects to turn a profit and to close another $6 million venture capital round by midyear.

While its technology appears to hold great promise, Civil Maps faces stiff competition in the self-driving car market, which is expected to reach $42 billion by 2025. Last year, GM acquired Cruise Automation's software and 40-person team for $1 billion. Other competitors such as Argo AI have been getting frequent investments and acquisition offers from old-guard automakers.

Even more daunting is the technical feat the company is attempting to accomplish. There are 4.2 million miles of road in the United States, according to the Department of Transportation, and mapping them all down to the centimeter isn't easy. "We've had dynamic maps in cars for years, but they're not very accurate," says Karl Brauner, publisher of Kelley Blue Book, referring to GPS devices and apps like Garmin, TomTom, and Waze. "Your car could be 10-30 feet away from where it thinks it is. But when you have a self-driving car, you want to be hyper accurate so you know you're on the right side of the road and between the lanes."

Civil Maps is trying to deliver the needed level of accuracy by crowdsourcing driving data from every car running its software. But rather than uploading everything from each vehicle (which can generate a gigabyte of information every second), the software uses compression along with algorithms to distinguish between what is necessary information (signage, traffic lights, people), and what isn't (buildings, billboards, birds). This geospatial data is then distributed to all Civil Maps-equipped-cars from the cloud as needed, Pattagunta says, adding that updates will eventually be available weekly. That means more recent maps than traditional GPS mapping software can provide. No more unforeseen construction throwing a wrench in your commute, and paired with Civil Maps' cognition system, no collisions, either.

Ultimately, Civil Maps aims to make people comfortable with the idea of giving up control over their vehicles. When there's no human driver, "the car has a responsibility to communicate what it knows and understands," says Puttagunta. "That way people can trust the car over time."