Keeping OSM fresh, accurate, and navigation-worthy at Lyft

Written by Brian Spencer, Rostom Zain El-Din, Yuliya Shustava, Anastasiya Prakopava, and Kiryl Yakimavets

Lyft’s mission to improve people’s lives with the world’s best transportation requires investing in the world’s best map. Nothing could be accomplished without first having a reliable groundwork of highways, buildings, and natural features through which to guide drivers and riders. We rely heavily on accurate and up-to-date map data — this is why Lyft Mapping is built on OpenStreetMap (OSM).

OSM is a global map database used by millions of people around the world, for tracking agricultural land use, disaster recovery, refugee response, academic research, and much more. After 19 years of growth, OSM is now commonly used by many companies to power applications like logistics platforms, social media, and gaming. OSM is now the biggest crowdsourced repository of human geospatial knowledge — and Lyft is proud to be a contributor since 2020.

Why Lyft Chooses OSM

When we first evaluated OSM, we found that it aligned with Lyft’s goals and needs thanks to three key findings:

High-quality road network

OSM already had a robust core road network across our major markets in North America. This meant that for basic needs like turn-by-turn navigation, the road geometry and connectivity were sufficient to get us started.

Large and active community

With more than 10 million registered users, OSM effectively harnesses the power of crowdsourced mapping. This community-driven approach ensures several benefits:

Fresh Maps — With peers and partners working on the same map, it stays fresh and accurate
Incremental Maintenance — Partially completed map layers mean that only incremental updates are needed, making it easier to maintain
Engaged Community — The OSM community is actively involved in defining tags and schemas, ensuring that the map evolves to meet the needs of its users
Collaborative Efforts — Many companies are working on similar map layers for their applications (e.g., gated communities), which means we can leverage shared knowledge and resources

Thriving OSM ecosystem

The open source nature of OSM combined with its large community make for a thriving and innovative ecosystem, contributing to two distinct advantages:

Tools and Resources — A plethora of tools and resources are available for map editing, analysis, and geospatial processing
Rapid Response — The ability to improve the map ourselves allows us to respond quickly to customer feedback and make necessary updates

Initial Challenges & Growth

Of course, working on a project with ten million of your closest mapping friends isn’t all peaches and cream. Lyft faced some hurdles during the initial phase of contributions to OSM:

Policy Interpretation — We were tasked with studying all possible OSM sources and guidelines related to our projects. However, not all elements of our projects are specifically addressed by guidelines or OSM best practices, requiring us to make interpretations based on local mapping peculiarities. The Lyft Team continues to mature its mapping skills through practice and communication with the community.
The complexity of changing OSM mapping rules — Lyft diligently adapts its projects to adhere to OSM guidelines. In some cases, however, we develop best practice updates that we believe are useful for the entire OSM community. But the process of clarifying or changing OSM guidelines requires active participation from all interested parties. It can be very time-consuming and does not always yield the expected results.
Lyft-Owned Evidence — OSM mappers initially questioned changesets where we used our own data as evidence. Now, all Lyft OSM profiles include source descriptions, and we provide necessary data upon request. This has reduced confusion and increased trust in our edits.

Over time, we learned how to best leverage OSM and its community for success in Lyft Mapping and quickly expanded our efforts. We began in 2020 with five Lyft mappers focused on identifying missing roads using our internal automated pipeline and adding or editing ‘lanes’ and ‘turn:lanes’ tags. These two projects laid the foundation for our maps, providing us with initial experience in detecting and resolving map issues, which subsequently increased the number of edits we made.

After over four years of working with OSM we now have 38 Lyft mappers contributing over 3K edits per week across six mapping features (turn restrictions, barriers & access, highway signs, road lanes, missing roads, and complex investigations — more details on the Lyft OSM wiki page).

Making a Difference

Lyft is proud to be contributing to an ecosystem used by so many. Editing the map is inspirational for us, especially when our expertise and unique data help resolve complex situations and contribute to critical map updates.

Unique sources for edits lead to a better, fresher map

In addition to all available open sources, Lyft is collecting ground truth and driver telemetry data to update maps with real-time road construction and environment changes.

The left screenshot is the freshest available Nearmap background (March 2024) of the intersection without a roundabout. However, Lyft driver telemetry data reveals that the traffic is organized with a roundabout, allowing us to apply the needed map update.

Lyft drivers provide an additional source of needed map updates as they report issues that occur during rides. Some reports lead to huge improvements at major intersections.

In the example above, the driver initially reported two missed roundabouts, but upon further investigation, we discovered that the area around I-80 and County Road F44 required extensive editing.

Ultimately, Lyft’s driver feedback and image collection programs enable rapid map updates for road changes, ensuring accurate navigation, improved customer satisfaction, and optimized routes for reduced fuel consumption and environmental impact.

OSM Community Engagement & Contributions

Lyft aims to be an engaged and collaborative member of the OSM community. We have taken on several initiatives and practices to stay connected with other mappers and contribute to the continued improvement of OSM mapping practices.

Community communication

Our team monitors several communication channels to maintain two-way discussions with OSM community members, the main one being public changeset discussions. Since 2020, we have engaged in over 265 changeset discussions. Most were related to clarifying the evidence that we used for specific map edits, but some discussions contained constructive feedback that helped us improve our curation approach.

Still other discussions contain positive feedback that highlight our efforts and motivate us to maintain high mapping quality levels. Engaging with the OSM community and receiving feedback improves our mapping expertise and fosters learning.

We also maintain communication through email (dct-osm(at)lyft.com), OSM forums and Slack channels, which are used mainly for clarifying information to launch new projects or updates. For example, we initiated two forum discussions (first, second) to resolve a disagreement about adding destination tags on motorways in the USA and found an appropriate way to add destinations in Canada. The discussions were aimed at finding a consensus on the best approach to tagging destinations, ensuring that the data would be useful and accurate for all users.

Community contributions

The Mapping Curation Team at Lyft created and maintains OSM training on GitHub. These guides help improve OSM editing accuracy for our internal mapping teams and are accessible to any new mappers outside of Lyft. OSM community members appreciate and share these guides on social media, promoting best mapping practices.

Impact

In the span of four years of mapping, the Lyft team has published over a half a million changesets with 5.5 million edits. Our contributions have included, but are not limited to:

135k changesets for adding new ways and correcting the geometry of existing ones, checking way directions, validating tags that define road drivability
142k lanes editing changesets to help drivers better navigate through interchanges and multi-lane intersections
67k edits based on destination signs to make guidance on motorways as useful as possible
37k turn restrictions adjustments including newly-added restrictions to ensure drivers are offered the best possible route without any illegal turns

All this impact is thanks to the OSM ecosystem, its incredible community, and a dedicated team of mappers from the Lyft Curation Team.

As we continue to grow our mapping efforts at Lyft, we will focus on improving road network tagging and geometry in OSM with a focus on adding new roads, turn restrictions, lane and destination data, road barriers and maintaining construction updates using both public and private data. We stay committed to open communication with the OSM community members. Together we will build the world’s best map to support the world’s best transportation.

Interested in learning more? Lyft’s own Jason Laska will be speaking in detail on this topic at OSM State of the Map US 2024. We’d love to see you there!