The @CITY‑AF project, carried out by Audi AG in cooperation with a consortium of research institutes and universities, aimed to develop and validate automated driving functions for urban streets and complex junctions. The work built on the foundations laid in the earlier @CITY programme, which had produced algorithms for environment perception, situation understanding, high‑precision mapping and precise vehicle localisation. In @CITY‑AF these algorithms were translated into operational driving functions that could be deployed in a production‑grade vehicle.
A central technical achievement was the integration of a multi‑sensor suite—including lidar, radar, cameras and high‑definition GPS/IMU—into a dedicated test vehicle. The sensor fusion architecture delivered a robust perception of the surrounding traffic environment, recognising vehicles, cyclists, pedestrians and static obstacles with a detection rate that exceeded 95 % in the test scenarios. The localisation system, based on a pre‑built map with centimetre‑level accuracy, maintained a lateral error below 0.15 m and a longitudinal error below 0.2 m during all test drives. These metrics were verified on the Aldenhoven Test Centre, where the vehicle performed a series of manoeuvres through roundabouts, signalised intersections and unsignalised junctions, as well as interactions with vulnerable road users at crosswalks.
The driving strategy module was designed to handle the full spectrum of urban driving tasks. It generated lane‑keeping and speed‑control commands that respected traffic rules and dynamic constraints, while also executing safe lane changes and yielding behaviours at intersections. The system’s decision logic was validated in simulation and on‑road tests, demonstrating a successful completion rate of 98 % for the predefined test cases. In addition, the interaction module enabled the vehicle to anticipate the intentions of pedestrians and cyclists, adjusting its trajectory accordingly. The module’s reaction time to a pedestrian step onto the road was measured at 0.45 s, well within the safety margins required for urban driving.
The project also produced a set of reusable software components and a validation framework that can be integrated into future Audi models. The modular design allows for rapid adaptation to different vehicle platforms and sensor configurations, facilitating the planned rollout of automated driving functions across Audi’s mid‑size and compact range.
Audi AG led the overall project management and the integration of the automated driving stack into the test vehicle. Partner organisations contributed specialised expertise: universities supplied research on perception algorithms and machine‑learning models; research institutes provided simulation environments and validation tools; and a technology supplier supplied the high‑definition mapping database. The consortium worked closely with the German Federal Ministry of Economic Affairs and Climate Action, which funded the initiative under grant number 19A18003C. The project ran from 1 July 2018 to 30 June 2022, with a cost‑neutral extension until 31 August 2022 to accommodate delays caused by the COVID‑19 pandemic. Throughout the four‑year period, all partners met the agreed milestones, culminating in a public presentation of the first prototype vehicle and a final report that documents the technical outcomes and the roadmap for commercial deployment.