The UNICARagil project, funded under the German federal programme “Microelectronics from Germany – Innovation Drivers of Digitalisation” (grant number 16EMO0286), ran from 1 February 2018 to 31 May 2023, a total of 64 months, with a 16‑month extension due to work‑time restrictions. The Technical University of Darmstadt (TU Darmstadt) led the work packages on modular assurance (AP 1.5) and motion control with safe stopping (AP 2.4). TU Darmstadt collaborated with the Institute for Mobile Robotics (iMAR) and other partners in the field of vehicle dynamics and satellite geodesy. The project’s goal was to create a disruptive modular architecture for autonomous agile vehicle concepts, focusing on the verification of driving capabilities and the safe operation of vehicles with four independently steerable wheels.
A central scientific contribution is the Behaviour‑Semantic Scenery Description (BSSD), a novel representation that captures lane‑level geometry and dynamic elements with semantic detail beyond current standards. Using a two‑stage methodology, the team derives lane‑accurate routes from the BSSD, identifies transitions between behavioural spaces, and formulates a comprehensive set of driving requirements. These requirements are then mapped to corresponding driving capabilities through specification categories that link physical scenery features—such as lane geometry or intersection geometry—to behavioural demands. The resulting catalogue of requirements and capabilities is applicable to any deployment area modelled with the BSSD, and has been instantiated for the entire street network of Darmstadt and for the demonstration site in Aldenhoven.
The capability‑based route planner builds on the open Lanelet2 map format. It performs a direct comparison between the derived requirements and the vehicle’s capabilities along a planned route, ensuring that autonomous operation is only authorised on segments where the vehicle’s capabilities suffice. For the final demonstration, a new mission specification format was introduced, containing route data and target states. Mission information is stored in the cloud, allowing both vehicle applications and the vehicles themselves to retrieve or modify mission parameters in real time.
The project’s scientific output is documented in a doctoral thesis by Moritz Lippert (submitted 29 November 2022, defended 17 January 2023) and in several peer‑reviewed publications. These works detail the derivation of behavioural attributes for the BSSD, the development of the BSSD for road networks, the extraction of route‑based behavioural requirements, and the capability‑based route planning framework. The thesis and papers provide quantitative performance data, such as the accuracy of requirement extraction and the computational efficiency of the route planner, though specific numerical values are not reproduced here.
In the motion‑control domain, AP 2.4 addressed three core modules: vehicle dynamics state estimation (FZS), vehicle dynamics and trajectory following control (FTR), and safe stopping (SiA). The FZS module fuses GNSS, inertial measurement unit, and odometry data, while the FTR module adapts established single‑steer control algorithms to the four‑wheel‑steer architecture of UNICARagil vehicles. The SiA module defines emergency‑stop manoeuvres for failures of essential services, a novel concept for agile vehicles that had no prior solutions at project start. Together, these modules form a closed‑loop control system essential for autonomous operation.
Overall, the project delivered a comprehensive framework that links detailed environmental semantics to vehicle‑specific capabilities, a capability‑based route planner that enforces safety constraints, and a modular control architecture for agile autonomous vehicles. The collaboration between TU Darmstadt, iMAR, and other partners, supported by the German federal funding programme, enabled the integration of advanced perception, planning, and control techniques into a coherent, verifiable vehicle architecture.