Impact of Network Latency on the Lateral Control in Automated Vehicle Marshalling

Automated Vehicle Marshalling (AVM) is the first functionally safe Level 4 automated driving system. It consists of the wireless control of unoccupied vehicles at low speed in well-defined environments, such as parking facilities or manufacturing plants. The driverless operation in an AVM system is achieved by transmitting control messages between connected vehicles and intelligent infrastructure. Similar to other wireless applications, network reliability poses a major challenge to ensuring safe automated driving. An AVM system must provide uninterrupted communication between the vehicle and the infrastructure at a stable frequency. However, wireless systems usually suffer from varying latencies and network disturbances. In this context, international organizations and automotive industry contributors have defined requirements specifying network performance, communication interfaces, and message formats for different AVM use cases. These requirements cover communication aspects without involving core automated driving functions, such as vehicle motion control, which are also decisive in ensuring the safety of the overall system. Therefore, studying communication factors in combination with vehicle motion control offers better interpretability of system capabilities. In this work, we investigate the trade-off between communication specifications and vehicle lateral control within an AVM framework implemented on a real vehicle. We aim to address the limitations that may arise under real-world AVM driving conditions. First, we revisit the current technical specifications to highlight the specific AVM messages relevant to vehicle lateral control. Then, we propose a testing framework by establishing communication with the test vehicle over a Wi‑Fi network using multiple access points deployed across an indoor parking facility and an outdoor test track. Thus, we obtain a quantitative analysis of network factors, such as latency, in different driving environments. In the next step, we present a model predictive control (MPC) approach that uses the AVM control messages to achieve robust vehicle lateral control. By evaluating the control performance under communication conditions, we assess the impact of network latency on vehicle lateral control. This work provides a baseline for exploring the limitations of AVM and deriving potential optimizations.