When Google Maps debuted in 2005 the technology amazed users, who could locate an address and even visit the spot using archived “Street View” captures. The next generation of digital maps for autonomous driving and augmented reality (AR) will require much more — for example that the states and positions of rendered real world objects be both precise and up-to-the-minute.
These new challenges and opportunities have prompted mapping companies to reimagine digital mapping technology.
Mapbox, a San Francisco-based digital mapping startup, released its Mapbox AR at the Mobile World Conference 2018 in Barcelona. The AR platform comprises a suite of tools, a low-level AR core kit, and a framework. Users can access interactive maps with animated dynamic directions on mobile devices.
Founded in 2011, Mapbox offers clients development tools for mapping their apps and website, including a featured Photoshop-like product, Studio, that enables developers to access Mapbox’s visualization layers and make maps with their own data. Mapbox business model focuses on visualizing for example high-density weather radar information; delivering business analytics; and logistics.
While Google is unquestionably a dominant player in the digital mapping industry, Mapbox is carving out a niche. Their Road Network collects anonymous data information from 300 million users of apps such as Airbnb, Instacart, Snap, and others with embedded Mapbox SDK. This data can extrapolate detailed patterns, for example, of real-time traffic and environmental status.
Dave Cole, a Mapbox founding member and VP of Business Operations and Strategy told Synced that data is the key to his company’s success. “We have a network now that’s collecting over 220 million miles of anonymous data every single day that goes back into the map and it gets better.”
Mapping for autonomous driving is a high-growth sector estimated to open up a US$20 billion market by 2050, according to Goldman Sachs. While breakthroughs in self-driving technologies have thus far come from sensors such as LiDars and Cameras, the important role of maps cannot be overlooked. In the future, maps may even become more important than sensors.
“For example, when driving on a multi-lane road, the onboard sensor may be unable to detect the road after a turn because of road-side obstructions. Once you have the lane-level positioning and real-time road updates enabled by HD maps, you can slow down and switch to a safe lane in advance to prevent an accident,” says Xudong Cao, Founder and CEO of China’s self-driving startup Momenta.ai.
Mapping companies are vying for the vanguard in autonomous driving integration. Dutch mapping firm TomTom is a leader in HD map production, and has attracted investments and partnerships deals from Ford, GM, Fiat Chrysler.
Mapbox made its initial move into autonomous driving in 2016 with Mapbox Drive — now Mapbox Automotive — a package of developer tools that enabled automakers to customize HD maps and turn-by-turn navigation. Developers can access API offered by Mapbox Automotive and build out specific embedded dash navigation, as well as 3D rendered visualization, to give drivers enhanced environmental context.
Mapbox Automotive has also built a Global Lane Network that aggregates sensor data collected by cameras, LiDars and GPS from other self-driving partners and distributes the information.
“Think about a lot of autonomous companies that have invested in computer vision or LiDar. They are generating all this data on the vehicles, but they need a network to actually distribute it back so as one vehicle collects, they can refine the map and make it available for general use for the whole fleet of vehicles,” says Cole.
To maintain a map that accurately keeps up with the world as it changes, Mapbox researchers are leveraging AI technologies for processing data and turning it into, for example, predictive traffic profiles based on density of people; and using Receptive Field Networks (RFNet) or You-Only-Look-Once (YOLO) algorithms for segmentation and object recognition.
Last year Mapbox purchased Mapdata, an AI mapping startup that uses deep neural networks to improve computer vision and AR. The Belarus-based company is now a satellite R&D team devoted to merging front-facing cameras and navigation.
Mapbox has a team of 350 and has thus far accumulated investments of US$227 million. As of February 2018, the number of registered developers on the Mapbox platform topped one million, a milestone for a technology platform.
While 60 percent of its business is with US developers, Mapbox is broadening its global operations, particularly in China. This month the former Head of Uber Business Development in Asia-Pacific (APAC) Andy Lee joined Mapbox to lead its APAC expansion.
Lee told Synced that because Mapbox is already partnered with Chinese companies such as Alibaba’s trip service Feizhu, it is in a position to help Western companies enter China, and its Chinese partners go global.
“What we have done is helpful for many companies who are trying to figure out how to enter China. Building maps in China is a challenge of being able to deliver experiences on a very large scale, not just like hundreds of thousands but sometimes hundreds of millions of users on a daily basis. There are very interesting engineering problems,” says Lee.
When Cole spoke to us he shared a story about his youthful hobby of sketching building and drawing maps. Coincidentally, the young Lee loved to carry paper maps his mother gave him and study streets “in a very nerdy way.”
“The experience has totally changed now because I have a smartphone with a map and travel or dining or lifestyle apps. My early childhood interest in maps is my career now, which is helping developers continue that journey, and that gets me really excited,” says Lee.
Journalist: Tony Peng| Editor: Michael Sarazen