Follow
Subscribe

Silicon Valley Company DeepMap Announces ‘RoadMemory' a Scalable Mapping Service for Autonomous Vehicles

Home > News > Content

【Summary】Mapping startup DeepMap Inc announced its new crowd-sourced mapping service for autonomous vehicles today called “RoadMemory”. It's designed to expand map coverage for autonomous vehicles more quickly and efficiently. RoadMemory will automatically build updated HD maps using crowdsourced data collected from vehicle sensors, such as cameras, radar and lidar.

Eric Walz    Jul 10, 2021 2:30 PM PT
Silicon Valley Company DeepMap Announces ‘RoadMemory' a Scalable Mapping Service for Autonomous Vehicles
DeepMap uses crowdsourced data to build HD Maps for autonomous vehicles. (Photo: DeepMap Inc.)

Silicon Valley mapping startup DeepMap Inc announced its new crowd-sourced mapping service for autonomous vehicles today called "RoadMemory". 

DeepMap, founded in 2016, is a developer of high-definition (HD) mapping and localization technology. The company is headquartered in Palo Alto, CA, with additional offices in Beijing and Guangzhou, China. 

DeepMap's RoadMemory enables automakers to accelerate the creation and deployment of large-scale digital maps using crowdsourced data collected from fleets of passenger vehicles and trucks. It is designed to expand map coverage for autonomous vehicles more quickly and efficiently.

RoadMemory will automatically build updated maps using crowdsourced data collected from vehicle sensors, such as cameras, radar and lidar. More importantly, RoadMemory is sensor-agnostic, allowing automakers to use perception sensors from different manufacturers, which provides a higher degree of flexibility.

"Building maps has been a large and expensive undertaking until now," said Mark Wheeler, co-founder and CTO of DeepMap. "Now it is possible to create the large-scale, fresh, accurate, open, and economical maps the transportation industry needs to advance to higher levels of autonomy."

DeepMap said its developed RoadMemory in response to growing demand from automakers seeking scaleable, high-performance, and economical mapping capabilities to support advanced driver assist system (ADAS) features, such as highway assist, intelligent braking, and traffic jam pilot. These types of Level-2 automated driving systems include Tesla's Autopilot and Super Cruise from General Motors. 

Although current Level-2 systems are not considered to be "fully self-driving", they are likely to be in widespread use in the auto industry in the near future to improve safety and make the driving experience more comfortable, especially on highways.

"RoadMemory addresses the immediate autonomy needs of near-term production vehicles while providing a natural roadmap for future higher levels of autonomy," said DeepMap CEO James Wu. "DeepMap is offering a future-proof 'best of both worlds' approach which leverages the sensor capabilities available in cars today while increasing the fidelity of maps over time as cars become more capable."

70c54cb6-b83a-4f0d-9f0c-d74cf0fe3998.jpeg

DeepMap co-founders Mark Wheeler (left) and James Wu. (Photo: Bloomberg)

RoadMemory is designed to compliment DeepMap's HDR (High-Definition Reference), which is a service for automakers and other companies developing hands-free Level 2+ automated driving systems for the auto industry. DeepMap says its HDR solves a critical piece of the puzzle for companies seeking to validate and improve crowd-sourced map data. 

"Automated driver assistance and Level 2+ through Level 3 capabilities have become a critical battleground for automakers," said Wheeler. "Camera and lidar sensors are starting to ramp up and proliferate for near-term production models. In tandem, automakers are evaluating and choosing their map source."

Before an autonomous vehicle can safely navigate, it needs a highly detailed map of its environment, in addition to knowing its exact location in the world. The detailed maps are an integral part of a self-driving vehicle's AI-powered brain and are just as important as camera and lidar data. The data to build the HD maps is collected from vehicle sensors while driving through the area. 

DeepMap fuses crowdsourced images from digital cameras, radar and 3D lidar data collected from a vehicle to create its maps. The HD maps are accurate down to roughly three centimeters.

HD Maps include details not found on ordinary maps like those used for turn-by-turn driving directions. The 3D maps include detailed lane markings, and the exact position of road signs and curbs. However, ensuring that a vehicle's maps are always up to date and available is a big challenge for automakers as well as for developers of self-driving vehicles. 

For example, a lane closure due to construction requires an updated map be pushed to all of the vehicles relying on it for navigation. DeepMap's technology ensures that updated maps are readily available in real time. The company also offers efficient map data storage which supports low latency communication between vehicles and the cloud.  

Before co-founding DeepMap, Wu led the engineering efforts building the serving infrastructure of Google Earth and also helped launch Apple Maps. He also served as principal architect for Baidu's Apollo, an open self-driving platform in China.

In April, DeepMap was named as one of the most innovative companies utilizing artificial intelligence (AI) by CB Insights, a National Science Foundation-backed big data company. Deepmap was selected from a pool of over 6,000 companies that are redefining industries.

DeepMap is also well funded. It's backers include Andreessen Horowitz, Goldman Sachs, NVIDIA, and Robert Bosch Venture Capital

Prev                  Next
Writer's other posts
Comments:
    Related Content