The UNICARagil project delivered a novel, modular vehicle architecture that combines a mechatronic construction kit, a fully service‑oriented software stack, and a functionally safe electrical‑electronic (E/E) platform. Central to the effort was the design and verification of a capability‑monitor framework that provides self‑perception for automated driving systems. The framework aggregates quality metrics from the vehicle’s subsystems and exposes them at runtime, enabling the vehicle to assess its own operational readiness against the defined operational design domain (ODD). Implementation of the framework as an Automotive Service‑Oriented Architecture (ASOA) service allowed it to run at the sensor‑frame rate, ensuring that the autonomous system can react promptly to environmental changes.

Performance measurements confirmed that the capability‑monitor framework meets real‑time constraints. Latency tests of synchronous and asynchronous communication channels showed that end‑to‑end delays remained below the critical thresholds required for safety‑related data exchange. Arbitration times for the distributed runtime environment increased with the number of active links but stayed within acceptable limits, demonstrating that the system can scale to larger networks without compromising determinism. Convolutional neural network (CNN) inference times were evaluated against varying L3 cache miss rates; even under high miss rates the execution times stayed within the real‑time budget, proving that the hardware‑software co‑design supports the computational demands of perception modules.

The hardware‑software platform itself was engineered for real‑time capability. A mixed‑criticality communication runtime environment was introduced to filter critical from non‑critical data streams, thereby reducing interference and preserving timing guarantees. The platform’s time‑specification model, based on Logical Execution Time (LET) zones, was validated on a distributed architecture that integrates on‑chip memory (OCM) and precision time protocol (PTP) synchronization. The resulting system achieved deterministic message delivery across the vehicle network, a prerequisite for functional safety certification.

In parallel, the consortium developed the autoELF vehicle, a privately owned, app‑controlled car designed for family use. The vehicle’s interior was conceived to accommodate diverse user groups, with modular seating and a “sitzkiste” concept that allows rapid reconfiguration. Software integration of interior services was achieved through the same ASOA framework, enabling seamless coordination between the vehicle’s physical and digital components. Extensive testing and troubleshooting phases verified that the vehicle meets safety and usability requirements, and the final prototype was demonstrated in a real‑world scenario.

Collaboration within UNICARagil involved the Technical University of Braunschweig and the Technical University of Darmstadt as lead partners, with additional industry and research partners contributing expertise in mechatronics, software architecture, and safety engineering. Work packages 1.4.4 and 2.5.1–2.5.8 focused on functional safety, self‑perception, and diagnostic capabilities, while 4.4.1–4.4.6 addressed the design, implementation, and testing of the autoELF vehicle. The project was funded by the German Federal Ministry of Education and Research (BMBF) under the UNICARagil programme, running from 2021 to 2024. The consortium’s integrated approach—combining modular hardware, service‑oriented software, and rigorous safety verification—demonstrates a scalable pathway toward next‑generation autonomous vehicles.