Introducing: The Fingerprint Base Map™ for Autonomous Vehicle Mapping and Localization

By | January 9, 2018

By Scott Harvey, Engineering

Today, as CES 2018 kicks off, Civil Maps officially announces something we’ve been working on and refining for quite a while: our Fingerprint Base Map™ technology. It is a critical component of the breakthrough system that we’ve developed to enable cognition for self-driving cars. The Fingerprint Base Map is a key layer in our HD Semantic Map. It enables autonomous vehicles to understand where they are in the world and what is around them. This real-time localization to our map is particularly helpful in complex driving environments. While traditional solutions are available, they are often derived from legacy mapping systems and merely adapted to the task of autonomous driving. They are usually not robust enough and are too expensive to incorporate into vehicles beyond the research and development stage. With this in mind, we’ve architected our system from the ground up with scalability and precision as goals. This one technology changes everything.

Our fingerprinting process tackles one of the hardest problems in deploying a workable mapping and localization solution: the development of 3D semantic base maps that are detailed, localization-ready, and upgradable, all while remaining compact enough to be stored within production cars. Additionally, we are able to localize vehicles in six degrees of freedom (6DoF) within 10 cm accuracy via maps that can easily be crowdsourced. Ours is a robust solution as well, performing reliably in challenging weather, at different driving speeds, and at any time of day.

For scalability, size definitely matters when it comes to creating and updating maps. It is really the downfall of other systems with regards to their autonomous vehicle development programs. To create and maintain very large base maps, each vehicle requires very expensive hardware and involves labor-intensive, inefficient processing, including the manual delivery of mapping data.

Multiply those costs by hundreds and thousands of vehicles. The problem becomes obvious. Of equal concern is that the round trip for the data is not just expensive — it’s also slow, error-prone, and it extends the intervals between map refresh cycles to weeks and months. Conversely, our system is edge-based; the bulk of the map data processing is done in real-time, in-vehicle. With most conventional mapping and localization systems, the base maps can’t be updated in a timely fashion and therefore quickly become obsolete. This is clearly unacceptable for autonomous vehicles that require current, accurate maps to safely transport passengers.

Civil Maps’ Fingerprint Base Map technology solves these problems. The compact size of our base maps offers game-changing advantages:

  • Small data footprint allows for minimal hardware requirements, storage, computing power, and energy consumption
  • Real-time, edge-based data processing can be handled in-vehicle with low-cost CPUs
  • New base maps can be downloaded to the vehicle, as-needed, via current 3G and 4G cellular systems, cached when reception is poor


The way our fingerprinting solution works is not that different from the technology Shazam — recently acquired by Apple — uses for song identification. Shazam creates a digital summary or an “audio fingerprint” of each recording in its database. When there is a song to be identified, Shazam records the user’s input and generates a new audio fingerprint for it that can be compared to those in its database in order to find a match.

In the case of Civil Maps, cars deployed with our technology stack drive around gathering and processing sensor data in-vehicle. Our algorithms transform those gigabytes of raw sensor data into lightweight, voxel-based fingerprints. These fingerprints enable cars to localize as they’re matched to other fingerprints in our map reference database. Similar to traditional, iterative closest point localization (ICP), there are at least two cars involved: The first car (“reference” car) that creates the Fingerprint Base Map and the second car (the “query” car) that uses the fingerprint data to find itself within the reference map. If you’re interested in learning more, we’ve made a video posted above to show how our base maps are created and how subsequent cars match to that reference base map with voxel-based fingerprints.

Fingerprint Base Map technology is just one piece of Civil Maps’ innovative plan to revolutionize the autonomous driving space. Throughout the year, we’ve been proving these techniques with our automotive OEM customers and mapping partners. The continuous feedback has helped us tremendously. This week in Vegas, we are showcasing voxel-based fingerprint localization and other solutions during live demos around the Las Vegas Convention Center. It should be a pretty exciting time and we are looking forward to connecting with many of you.

If you are considering a scalable mapping and localization solution for your autonomous vehicle program and you’d like to see a demo, contact us at or stop by Booth# 9317 in the North Hall.

Read the original article

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.