Qualcomm powers TomTom's high-definition mapping
Qualcomm Technologies, Inc. is working with TomTom to accelerate autonomous driving technology by using the Qualcomm Drive Data Platform for high-definition (HD) map crowdsourcing. The Qualcomm platform collects and analyzes data from different vehicle sensors to determine vehicle location, monitor and learn driving patterns, perceive vehicle surroundings, and share this perception reliably and accurately outside the vehicle.
TomTom's HD Map, including RoadDNA, is a highly accurate, digital map-based product that assists automated vehicles to precisely locate themselves on the road and help determine which way to maneuver, even when traveling at high speeds.
Traditional development of maps requires deploying dedicated fleets of vehicles that are equipped with professional-grade sensors to collect location, raw imagery, Lidar, and other data, which is then transferred, stored, and processed in data centers. Now that cars are increasingly connected and equipped with a range of sensors, new and complimentary approaches such as TomTom’s have become possible.
Using the precise positioning, on-device machine learning, and heterogeneous compute and connectivity capabilities of the Qualcomm platform, which features the Snapdragon 820Am automotive processor, TomTom and Qualcomm Technologies aim to facilitate adding an improved, scalable, and cost-efficient crowdsourcing approach to the mix of sources for HD map making. The new concept is designed to allow massive numbers of connected cars to see and understand their environment, traffic, and road conditions, and support real-time input for map and road condition updates.
"Feature-rich, highly accurate, and frequently updated HD maps are critical to support some of the most advanced applications envisioned in the automotive industry, especially for autonomous driving," said Willem Strijbosch, head of autonomous driving, TomTom. "We are building the cloud-based platform to make and maintain HD maps using a range of input sources, including crowdsourced data from swarms of intelligent connected vehicles."
Qualcomm Technologies believes that an affordable and easy-to-integrate mapping solution for autonomous vehicles is realizable. Machine-learning algorithms running on the Snapdragon 820Am processor are optimized using the a neural processing engine toolkit to identify features such as traffic signs and lane markers, then fuse this information with precise positioning to generate feature-rich, highly accurate, and lightweight map updates. This enables cars on the road to become live sensors that automatically update navigation services with real-time road conditions using the 820Am’s integrated X12 LTE modem. Object identification and classification processing runs on the heterogeneous cores of the Snapdragon processor, avoiding the need and cost of sending raw video over the network. The process is designed to further optimize over multiple iterations, running locally on the platform, and then upload to the cloud for crowdsourcing. The real-time HD map can be constantly improved with more highly accurate location information for safer and advanced driving.