State of the art automated valet parking systems (AVPS) use infrastructure-based sensors in combination with recent vehicle technologies. The systems are mostly designed to work in mixed traffic environments. The infrastructure-based sensors are therefore necessary to ensure safety of the whole AVPS. This leads to limited penetration of AVPS, as the infrastructure must be additionally equipped with costly technology. The objective of the presented research is therefore to develop a system architecture for a cost-effective AVPS, which profits particularly from already in-vehicle installed sensors and minimizes necessary investment to the infrastructure system. One group of published system architectures in the low-cost range mostly provide a SLAM approach. However, such an approach is not suitable for commercial deployment of AVPS due to the lack of comfort as a driver needs to drive across the parking facility before being able to use the system. Other authors use sensor systems in a configuration not usual for production vehicles, again limiting feasibility of a commercial implementation. The system architecture proposed in this paper relates to low-cost sensor configurations of current production vehicles. To eliminate the need for SLAM algorithms, the proposed architecture relies on a communication unit within the infrastructure to provide all necessary information about the parking facility. Primarily, the system requirements must be defined. Standards, recommendations, and specifications of already implemented systems serve as a point of reference for understanding necessary design constraints. Furthermore, the stakeholder needs are analyzed, and the overall system is decomposed into subsystems and modules to be able to describe the requirements completely in a specification sheet. Based on the system requirements, a system architecture is derived. For this purpose, state diagrams are first outlined to describe the functionality of the AVPS. An input-process-output (IPO) model then specifies which hardware and software elements are essential and how they interact with each other to be able to realize the defined requirements and functions. Finally, a hazard analysis identifies potential risks, which are then reduced by appropriate contra-measures. This research results in a comprehensive system architecture including a specification for a safe and cost-effective AVPS. The defined requirements consider currently published standards and recommendations and therefore also contain the expected as well as required functions of a commercial AVPS. Furthermore, the system has been developed to meet the recommended requirements of the German Federal Motor Transport Authority (ger. Kraftfahrt-Bundesamt, KBA) to get an approval for deployment in Germany. Based on the set of requirements, a detailed, manufacturer-independent system architecture was developed, which contains only requirements on hardware and software modules. This approach allows universal deployment of the proposed architecture of a safe and cost-effective AVPS. To validate the architecture, the open research platform for automated driving ANTON was selected for the deployment. ANTON is based on an open-source automated driving stack. The system architecture was implemented to a simulation environment to test the implementation before its deployment to the real vehicle.
Mr. Ömer Dönmez, Research Assistant, Technische Hochschule Ingolstadt