Transforming the HD map update process with crowdsourced road data
Note: Woven Planet became Woven by Toyota on April 1, 2023.
Few people doubt that automated driving is the future of transportation. It’s likely that the next generation will find it hard to believe that highways were once full of speeding vehicles controlled by humans, separated from each other only by paint and driver concentration. But as every engineer knows, despite significant progress there are still formidable technical challenges to overcome before automated driving can become a reality.
One of those challenges is road lane mapping. Or more specifically, road lane map updating. Automated driving and Advanced Driver Assistance Systems (ADAS) rely heavily on accurate high definition (HD) map data, and while it’s relatively easy to deploy survey vehicles to generate the initial data, the vehicles are highly specialized and complicated to operate, making it time-consuming to maintain and regularly update the maps. Road markings and structures change position after roadworks or construction, and these changes have to be reflected quickly into the HD maps. How can we do this in an efficient and cost-effective way?
Hitchhiking on everyday traffic
At Woven Alpha, Inc. (Woven Alpha), a group company of Woven Planet Holdings, Inc. (Woven Planet), our engineers are developing a solution we call the Automated Mapping Platform (AMP).
AMP is a type of crowdsourcing for lane information, based on an open platform that utilizes anonymized data from ordinary vehicles going about their regular business on public roads. Instead of using special-purpose survey vehicles traveling on pre-determined routes with the sole purpose of generating road map data, AMP hitchhikes on everyday traffic to detect road changes and generate new data that can rapidly and frequently update the HD maps at the heart of automated driving.
HD maps by DMP/Ushr
Dynamic Map Platform Co., Ltd. (DMP) and its U.S. subsidiary Ushr Inc. (Ushr) use ground-truth precision and automotive grade validation to create highly reliable maps that enable safe and confident hands-free driving assistance. Using their geospatial mapping expertise, they developed the world’s first centimeter-level, automotive grade HD map. DMP/Ushr have developed an affordable product suitable for market models, and with five years of production experience they are currently the world’s only market-tested HD mapmaker. In addition, they announced coverage expansion from highway roads to surface roads in April, 2021.
Idea validated in initial 2020 testing
The project has been in development for a couple of years, and in April 2020 we took it to the road with a single-vehicle proof of concept (“PoC”) trial that showed our AMP could be effectively used for change detection of HD maps for highway roads.
In this PoC, through joint testing with DMP, Woven Planet Group’s predecessor Toyota Research Institute — Advanced Development, Inc. (TRI-AD) successfully demonstrated that under everyday conditions, ordinary vehicles can be a source for updating HD maps with lane feature changes, using data from standard dashcams. The next step in the development process is to test the data capture process using cameras embedded in commercially available vehicles — foreshadowing our ultimate goal of crowdsourcing map data to the entire road user fleet.
Advanced PoC testing using cameras of commercially available vehicles
Starting May 2021, Woven Alpha’s Automated Mapping Team will partner with DMP to conduct the next PoC. The key difference from the initial PoC in the previous year will be our use of commercially available vehicles — cars that people drive off the showroom floor — as a data source.
This test will evaluate the use of AMP for highway road change detection for HD map-updates, via cameras embedded in commercially available production vehicles. AMP will use anonymized probe data (PD) collected from these vehicles to detect change points on highway roads. The change information will then be sent to DMP, who will update their HD map accordingly. If successful, this will unlock the door to efficient and automated mapping of road and lane change points — and bring us all closer to the world of automated driving.
Targeting commercial operations of highway road change detection
Our ultimate goal of this PoC is to make it possible to deliver consistently up-to-date high-definition road and lane network maps, starting with highway roads, based on a map update system that combines efficiency, accuracy and cost-effectiveness.
Successful completion of stage two PoC testing will be a significant milestone in our program. As more ADAS systems equipped with the DMP HD map data are launched in Japan and the US, we will continue working with DMP to develop and evolve our change-point detection capabilities in preparation for commercial operations covering Japan highway roads, followed by potential expansion to Japan surface roads and to roads in the US.
Creating open platform using crowdsourced data
AMP will be an open platform based on a contribution model. Ordinary vehicles with embedded cameras contribute anonymized sensor data to the platform to maintain up-to-date maps, and any developer will have easy, open and sustainable access to the HD map data.
The vehicles will be installed with embedded cameras from the factory. This means that as more vehicles are sold with this system, data will become increasingly abundant, paving the way for the massive expansion of map coverage needed to enable automated driving systems to navigate us accurately and safely along the world’s roads.
Since coming together under our new Woven Planet Group structure, we’ve accelerated toward our vision for automated mobility. It’s an exciting time to be engineering the future, and we look forward to sharing our progress via this blog as we develop and test our unique mobility solutions.
Join us! 👋
We are actively hiring exceptional talent for AMP. If you are interested in working with us in our brand-new office in Nihonbashi, Tokyo, please have a look at our current open positions!
