Frank Diermeyer

Gemb Kaljavesi, Xiyan Su, Frank Diermeyer

Online corner case detection is crucial for ensuring safety in autonomous driving vehicles. Current autonomous driving approaches can be categorized into modular approaches and end-to-end approaches. To leverage the advantages of both, we propose a method for online corner case detection that integrates an end-to-end approach into a modular system. The modular system takes over the primary driving task and the end-to-end network runs in parallel as a secondary one, the disagreement between the systems is then used for corner case detection. We implement this method on a real vehicle and evaluate it qualitatively. Our results demonstrate that end-to-end networks, known for their superior situational awareness, as secondary driving systems, can effectively contribute to corner case detection. These findings suggest that such an approach holds potential for enhancing the safety of autonomous vehicles.

Ines Trautmannsheimer, Ahmed Azab, Frank Diermeyer

Teleoperation promises to extend the operational envelope of automated vehicles, yet it critically depends on network latency and video quality. We report a fixed-base driving-simulator study (N=25) with a 2x2 manipulation of added latency (100/300 ms) and bitrate (500/2000 kbit/s), plus a best-case baseline (0 ms added, 9000 kbit/s). We measured effective glass-to-glass (G2G) latency per condition (baseline approx. 413 ms; effective totals approx. 500-700 ms) and verified stable framerate and encoder settings. Multimodal measures covered performance (speed, steering reversals, crashes), oculomotor behavior (blink rate, fixation duration), physiology (RR interval, heart rate, skin conductance), and subjective workload. Latency and bitrate each increased operator load and modestly affected performance. Physiological measures (heart rate, RR interval) exhibited sub-additive interactions, whereas performance and oculomotor interactions were small or non-significant. Equivalence tests showed that 300 ms with 2000 kbit/s was velocity-equivalent to best-case (SESOI +/- 2 km/h), while 300 ms with 500 kbit/s was not. We argue that latency and video quality should be treated as largely independent design levers, and that physiology-aware adaptation can anticipate overload before safety is compromised.

Ines Trautmannsheimer, Richard Grauberger, Frank Diermeyer

Automated driving has made remarkable progress, yet situations still arise where human intervention is necessary. Teleoperation provides a scalable solution to address such cases, enabling remote operators to support vehicles without being physically present. In this context, video transmission forms the operator's primary source of situational awareness, making video quality a decisive factor for both safety and task performance. In an online study, participants rated compressed video sequences from the Zenseact Dataset and provided subjective quality ratings. These ratings were then used to retrain the Video Multi-Method Assessment Fusion (VMAF) model, yielding an adapted variant tailored to teleoperation. The retrained model demonstrated improved alignment with human ratings compared to the original 4K VMAF. In particular, RMSE decreased from 10.36 to 8.83, and MAD from 8.71 to 6.38, corresponding to improvements of 15% and 27%, respectively. These results highlight that incorporating domain-specific data can enhance the predictive power of established quality metrics in safety-critical applications. At the same time, Outlier cases emerged in which videos received high objective scores despite noticeable degradations in regions critical for the driving task.

Nijinshan Karunainayagam, Nils Gehrke, Frank Diermeyer

Residual risk metrics have recently been introduced to assess the safety implications of automated driving systems. Existing approaches typically assume a deterministic ego pose and concentrate mainly on perception errors related to surrounding objects and latency effects. In practice, however, automated vehicles operate under considerable localization uncertainty, especially in complex urban settings and in adverse weather conditions. This work extends the spatial residual risk formulation to the belief space by explicitly modeling ego pose uncertainty as a Gaussian distribution. Residual risk is reformulated as the expected degradation-induced risk over the ego pose belief distribution. Within a particle-based risk estimation framework, localization uncertainty is incorporated into the computation of collision probabilities through covariance fusion of ego and object uncertainties.

Andreas Schimpe, Johannes Feiler, Simon Hoffmann et al.

Teleoperation allows a human operator to remotely interact with and control a mobile robot in a dangerous or inaccessible area. Besides well-known applications such as space exploration or search and rescue operations, the application of teleoperation in the area of automated driving, i.e., teleoperated driving (ToD), is becoming more popular. Instead of an in-vehicle human fallback driver, a remote operator can connect to the vehicle using cellular networks and resolve situations that are beyond the automated vehicle (AV)'s operational design domain. Teleoperation of AVs, and unmanned ground vehicles in general, introduces different problems, which are the focus of ongoing research. This paper presents an open source ToD software stack, which was developed for the purpose of carrying out this research. As shown in three demonstrations, the software stack can be deployed with minor overheads to control various vehicle systems remotely.

Andreas Schimpe, Simon Hoffmann, Frank Diermeyer

Vehicles with autonomous driving capabilities are present on public streets. However, edge cases remain that still require a human in-vehicle driver. Assuming the vehicle manages to come to a safe state in an automated fashion, teleoperated driving technology enables a human to resolve the situation remotely by a control interface connected via a mobile network. While this is a promising solution, it also introduces technical challenges, one of them being the necessity to transmit video data of multiple cameras from the vehicle to the human operator. In this paper, an adaptive video streaming framework specifically designed for teleoperated vehicles is proposed and demonstrated. The framework enables automatic reconfiguration of the video streams of the multi-camera system at runtime. Predictions of variable transmission service quality are taken into account. With the objective to improve visual quality, the framework uses so-called rate-quality models to dynamically allocate bitrates and select resolution scaling factors. Results from deploying the proposed framework on an actual teleoperated driving system are presented.

Thomas Ponn, Matthias Breitfuß, Xiao Yu et al.

For a successful market launch of automated vehicles (AVs), proof of their safety is essential. Due to the open parameter space, an infinite number of traffic situations can occur, which makes the proof of safety an unsolved problem. With the so-called scenario-based approach, all relevant test scenarios must be identified. This paper introduces an approach that finds particularly challenging scenarios from real driving data (\RDDwo) and assesses their difficulty using a novel metric. Starting from the highD data, scenarios are extracted using a hierarchical clustering approach and then assigned to one of nine pre-defined functional scenarios using rule-based classification. The special feature of the subsequent evaluation of the concrete scenarios is that it is independent of the performance of the test vehicle and therefore valid for all AVs. Previous evaluation metrics are often based on the criticality of the scenario, which is, however, dependent on the behavior of the test vehicle and is therefore only conditionally suitable for finding "good" test cases in advance. The results show that with this new approach a reduced number of particularly challenging test scenarios can be derived.

Thomas Ponn, Thomas Lanz, Frank Diermeyer

For the offline safety assessment of automated vehicles, the most challenging and critical scenarios must be identified efficiently. Therefore, we present a new approach to define challenging scenarios based on a sensor setup model of the ego-vehicle. First, a static optimal approaching path of a road user to the ego-vehicle is calculated using an A* algorithm. We consider a poor perception of the road user by the automated vehicle as optimal, because we want to define scenarios that are as critical as possible. The path is then transferred to a dynamic scenario, where the trajectory of the road user and the road layout are determined. The result is an optimal road geometry, so that the ego-vehicle can perceive an approaching object as poorly as possible. The focus of our work is on the highway as the Operational Design Domain (ODD).

Thomas Ponn, Fabian Müller, Frank Diermeyer

The objective of this paper is to propose a systematic analysis of the sensor coverage of automated vehicles. Due to an unlimited number of possible traffic situations, a selection of scenarios to be tested must be applied in the safety assessment of automated vehicles. This paper describes how phenomenological sensor models can be used to identify system-specific relevant scenarios. In automated driving, the following sensors are predominantly used: camera, ultrasonic, \radar and \lidarohne. Based on the literature, phenomenological models have been developed for the four sensor types, which take into account phenomena such as environmental influences, sensor properties and the type of object to be detected. These phenomenological models have a significantly higher reliability than simple ideal sensor models and require lower computing costs than realistic physical sensor models, which represents an optimal compromise for systematic investigations of sensor coverage. The simulations showed significant differences between different system configurations and thus support the system-specific selection of relevant scenarios for the safety assessment of automated vehicles.

Andreas Schimpe, Frank Diermeyer

Autonomous driving is among the most promising of upcoming traffic safety technologies. Prototypes of autonomous vehicles are already being tested on public streets today. However, while current prototypes prove the feasibility of truly driverless cars, edge cases remain which necessitate falling back on human operators. Teleoperated driving is one solution that would allow a human to remotely control a vehicle via mobile radio networks. Removing in-vehicle drivers would thus allow current autonomous technologies to further progress towards becoming genuinely driverless systems. This paper proposes a new model predictive steering control scheme, specifically designed for semi-autonomous, teleoperated road vehicles. The controller is capable of receiving teleoperator steering commands and, in the case of potential collisions, automatically correcting these commands. Collision avoidance is incorporated into the design using potential fields. A term in the cost function facilitates natural maneuvers, and constraints on the maximum potential keep the vehicle at safe distances from obstacles. This paper also proposes the use of high-order ellipses as a method to accurately model rectangular obstacles in tight driving scenarios. Simulation results support the effectiveness of the proposed approach.