The SafeADArchitect project, led by Intel Deutschland GmbH, ran from 1 October 2020 to 30 September 2023 and was carried out by a consortium of six partners: Intel Deutschland, Schaeffler Technologies AG & Co. KG, Lake Fusion Technologies GmbH, ANavS GmbH, the Forschungszentrum Informatik (FZI) research centre, and the Institute for Vehicle System Technology (KIT‑FAST) at the Karlsruhe Institute of Technology. The consortium was funded through a German federal research programme aimed at advancing automated driving safety. Intel served as consortium leader, coordinating project management, safety architecture design, and integration on the demonstrator vehicle, while Schaeffler supplied expertise on platform fault detection, Lake Fusion focused on perception and data fusion, ANavS developed visual‑SLAM modules, FZI contributed research on reliability and data‑generation methods, and KIT‑FAST provided vehicle‑system‑integration support.
The technical programme was organised into five work packages. AP0 established project governance and a risk‑aware management framework. AP1 produced a detailed requirements specification that linked functional safety goals to system‑level safety cases, drawing on ISO 26262 and ISO 21448 standards. AP2 created a comprehensive simulation environment for the complete system, enabling virtual testing of perception, planning, and control modules before hardware deployment. AP3 delivered a new system architecture that modularised perception, localisation, state monitoring, and safety‑critical decision making, and implemented it in a software stack that could be deployed on the demonstrator vehicle. AP4 integrated the full system onto the vehicle and performed extensive real‑world testing, while AP5 carried out a final evaluation against the original objectives, documenting performance, safety metrics, and lessons learned.
Key scientific results include the development of a LiDAR‑based SLAM module that builds on the Cartographer framework, and a visual‑SLAM system that incorporates inertial data for robust mapping. The project compared GPS‑only localisation with the combined GPS‑LiDAR approach, demonstrating increased resilience to signal loss and environmental changes. State monitoring of the vehicle platform was achieved by adapting methods from the SmartLoad5 project, which provided representative datasets and fault‑detection algorithms for mechatronic suspension components. Functional safety analysis confirmed compliance with ISO 26262, and the safety architecture was formally verified against the safety case derived in AP1. The demonstrator vehicle tests showed that the integrated system maintained safe operation under a range of driving scenarios, with localisation error reduced to sub‑meter levels and fault‑detection latency below 50 ms.
Collaboration across the consortium was tightly coordinated through regular steering‑committee meetings and shared development platforms. Intel’s role as consortium leader ensured that safety requirements were consistently translated into design decisions, while Schaeffler’s expertise on platform faults enabled early detection of potential failure modes. Lake Fusion’s perception algorithms were validated in the simulation environment before being deployed on the vehicle, and ANavS’s visual‑SLAM module was integrated into the overall architecture to provide redundancy. FZI’s research on data‑generation and reliability analysis underpinned the state‑monitoring component, and KIT‑FAST’s vehicle‑system integration expertise facilitated the seamless deployment of the software stack on the demonstrator chassis.
In summary, SafeADArchitect produced a validated safety architecture for automated driving, a robust localisation and state‑monitoring framework, and a demonstrator vehicle that achieved the predefined safety and performance targets. The project’s outcomes provide a solid foundation for future development of highly automated vehicles and demonstrate the effectiveness of a multidisciplinary, consortium‑based approach to automotive safety research.