DeepMap Plans to Teach Cars to Drive

Autonomous Vehicle Fever Inspires Another Start-Up

Recent years have seen a series of collaborations between the automotive and technology industries.

Mapping start-up DeepMap uses the “AI” top-level domain for its Web site, a clear suggestion that it sees maps as an essential ingredient to smart cars that can drive themselves. DeepMap, founded by former Google and Apple employees, just received $25 million in a funding round led by Accel, and there could be more to come.

Most car-makers around the world expect disruption to hit their market as autonomous driving approaches. But self-driving cars are essentially robots rather than tools and they require a special format of data. Navigation software such as Google Maps and Here WeGo were originally designed to interact with people and are being extended to provide layers of data suitable for machine use. But the next generation of mapping software will specialise to interact with machines rather than humans.

DeepMap aims to create algorithms and collect data to develop software that serves as the eyes and the brain needed to make cars autonomous. Its software works with sensors on a car and feeds the necessary information to give the vehicle updates on the environment, telling it exactly where it is in relation to its surroundings. The software also shares this information with other vehicles in the area which use the DeepMap platform.

There’s no doubt that a level of artificial intelligence will be needed to enable smart cars capable of dealing with unexpected road conditions and making real-world, split-second decisions. As autonomous vehicles manoeuvre through traffic and hazardous conditions, maps become a knowledge pool for decision-making and could become the basis of skills taken in reaction to changing road conditions.

DeepMap fuses images from digital cameras with data collected using lidar to build detailed 3D maps. Like other similar solutions, its platform can identify elements such as street signs, billboards and the height of a kerb. The 3D maps complement sensors on a vehicle by giving it a detailed awareness of the environment beyond the car’s view. DeepMap plans to license its map-building software to car-makers and technology companies interested in developing self-driving vehicles.

DeepMap isn’t alone in this area and faces steep competition from major technology firms including Alphabet, Apple, Intel, TomTom and Uber. These companies have the necessary navigation talent, experience, tons of data and industry contacts to push through their own intelligent mapping strategies.

Intel, which is a partial owner of mapping company Here, acquired computer vision firm Mobileye earlier in 2017. The market for autonomous cars has become an important part of Intel’s strategy, which plans to supply silicon and software to car-makers. Similarly, in 2016 General Motors snapped up Cruise Automation, and Ford recently committed to buying a majority stake in Argo AI.

Fully autonomous driving is still years away, but most players recognise that to realise the technology’s potential it’s crucial that the automotive and technology industries come together. This has been driving partnerships and acquisitions, as well as investments and cross-industry collaborations.

DeepMap’s move to make maps for machines rather than humans is a pure autonomous play. There’s no doubt that the company is getting significant attention, as navigation talent and vision become vital assets for the automotive industry.