Many AI systems, especially object and segmentation algorithms, have a high demand for good datasets. These datasets are also crucial for the development of automated and connected vehicles, but today there are few open/free datasets for the automotive industry. These datasets lack equivalent HD maps, repeatability and a closed, controlled environment. Therefore, VAMLAV – Validation of Mapping and Localization for Autonomous Vehicles – will create a new dataset which includes changes in the environment, a corresponding HD map and repeatability.
The dataset to be collected at the AstaZero test center will be used for validation and testing of map generation technologies and localization/positioning systems, using Zenuity vehicles equipped with various sensors. VAMLAV will also enable further research by making the dataset free and open to interested parties, by having AI Innovation of Sweden handle distribution. The transparent VAMLAV dataset enables other research areas, such as track planning, detection of environmental changes (weather, moving objects) and facilitates crowdsourcing etc. This provides an update and optimization of the maps currently held by Mapillary.
After project completion, VAMLAV aims at enabling other parties to contribute to the dataset at AstaZero’s Rural Road test track, thereby promoting research of crowdsourced HD map generation and the ability to share datasets within a specific industry.
VAMLAV will develop an HD map of Rural Road at AstaZero using the existing measurement infrastructure that is specifically installed at AstaZero for measuring positions and by designing and anchor points provided by RISE. VAMLAV also collects data from specific sensors for the dataset. The dataset and corresponding HD maps will then be used to generate new/other HD maps (image-based HD map) and help validate and test HD mapping generation and localization algorithms.
Project partners: AstaZero (coordinator), RISE, Zenuity, Mapillary, AI innovation of Sweden.
Project period: 1 Oct 2019 – 30 Sep 2021
Project budget: MSEK 11.5 of which 50% from Vinnova’s FFI program under contract no 2019-03097.