-
A Prototypical Expert-Driven Approach Towards Capability-Based Monitoring of Automated Driving Systems
Authors:
Richard Schubert,
Cedrik Kaufmann,
Marcus Nolte,
Markus Maurer
Abstract:
Supervising the safe operation of automated vehicles is a key requirement in order to unleash their full potential in future transportation systems. In particular, previous publications have argued that SAE Level 4 vehicles should be aware of their capabilities at runtime to make appropriate behavioral decisions. In this paper, we present a framework that enables the implementation of an online ca…
▽ More
Supervising the safe operation of automated vehicles is a key requirement in order to unleash their full potential in future transportation systems. In particular, previous publications have argued that SAE Level 4 vehicles should be aware of their capabilities at runtime to make appropriate behavioral decisions. In this paper, we present a framework that enables the implementation of an online capability monitor. We derive a graphical system model that captures the relationships between the quality of system elements across different architectural views. In an expert-driven approach, we parameterize Bayesian Networks based on this structure using Fuzzy Logic. Using the online monitor, we infer the quality of the system's capabilities based on technical measurements acquired at runtime. Our approach is demonstrated in the context of the UNICAR.agil research project in an urban example scenario.
△ Less
Submitted 25 April, 2024;
originally announced April 2024.
-
Conformal Prediction of Motion Control Performance for an Automated Vehicle in Presence of Actuator Degradations and Failures
Authors:
Richard Schubert,
Marvin Loba,
Jasper Sünnemann,
Torben Stolte,
Markus Maurer
Abstract:
Automated driving systems require monitoring mechanisms to ensure safe operation, especially if system components degrade or fail. Their runtime self-representation plays a key role as it provides a-priori knowledge about the system's capabilities and limitations. In this paper, we propose a data-driven approach for deriving such a self-representation model for the motion controller of an automate…
▽ More
Automated driving systems require monitoring mechanisms to ensure safe operation, especially if system components degrade or fail. Their runtime self-representation plays a key role as it provides a-priori knowledge about the system's capabilities and limitations. In this paper, we propose a data-driven approach for deriving such a self-representation model for the motion controller of an automated vehicle. A conformalized prediction model is learned and allows estimating how operational conditions as well as potential degradations and failures of the vehicle's actuators impact motion control performance. During runtime behavior generation, our predictor can provide a heuristic for determining the admissible action space.
△ Less
Submitted 25 April, 2024;
originally announced April 2024.
-
ODD-Centric Contextual Sensitivity Analysis Applied To A Non-Linear Vehicle Dynamics Model
Authors:
Richard Schubert,
Marcus Nolte,
Arnaud de La Fortelle,
Markus Maurer
Abstract:
Advanced driving functions, for assistance or full automation, require strong guarantees to be deployed. This means that such functions may not be available all the time, like now commercially available SAE Level 3 modes that are made available only on some roads and at law speeds. The specification of such restriction is described technically in the Operational Design Domain (ODD) which is a fund…
▽ More
Advanced driving functions, for assistance or full automation, require strong guarantees to be deployed. This means that such functions may not be available all the time, like now commercially available SAE Level 3 modes that are made available only on some roads and at law speeds. The specification of such restriction is described technically in the Operational Design Domain (ODD) which is a fundamental concept for the design of automated driving systems (ADS). In this work, we focus on the example of trajectory planning and control which are crucial functions for SAE level 4+ vehicles and often rely on model-based methods. Hence, the quality of the underlying models has to be evaluated with respect to the ODD. Mathematical analyses such as uncertainty and sensitivity analysis support the quantitative assessment of model quality in general. In this paper, we present a new approach to assess the quality of vehicle dynamics models using an ODD-centric sensitivity analysis. The sensitivity analysis framework is implemented for a 10-DoF nonlinear double-track vehicle dynamics model used inside a model-predictive trajectory controller. The model sensitivity is evaluated with respect to given ODD and maneuver parameters. Based on the results, ODD-compliant behavior generation strategies with the goal of minimizing model sensitivity are outlined.
△ Less
Submitted 13 April, 2023; v1 submitted 9 February, 2023;
originally announced February 2023.
-
Tackling the Class Imbalance Problem of Deep Learning Based Head and Neck Organ Segmentation
Authors:
Elias Tappeiner,
Martin Welk,
Rainer Schubert
Abstract:
The segmentation of organs at risk (OAR) is a required precondition for the cancer treatment with image guided radiation therapy. The automation of the segmentation task is therefore of high clinical relevance. Deep Learning (DL) based medical image segmentation is currently the most successful approach, but suffers from the over-presence of the background class and the anatomically given organ si…
▽ More
The segmentation of organs at risk (OAR) is a required precondition for the cancer treatment with image guided radiation therapy. The automation of the segmentation task is therefore of high clinical relevance. Deep Learning (DL) based medical image segmentation is currently the most successful approach, but suffers from the over-presence of the background class and the anatomically given organ size difference, which is most severe in the head and neck (HAN) area. To tackle the HAN area specific class imbalance problem we first optimize the patch-size of the currently best performing general purpose segmentation framework, the nnU-Net, based on the introduced class imbalance measurement, and second, introduce the class adaptive Dice loss to further compensate for the highly imbalanced setting. Both the patch-size and the loss function are parameters with direct influence on the class imbalance and their optimization leads to a 3\% increase of the Dice score and 22% reduction of the 95% Hausdorff distance compared to the baseline, finally reaching $0.8\pm0.15$ and $3.17\pm1.7$ mm for the segmentation of seven HAN organs using a single and simple neural network. The patch-size optimization and the class adaptive Dice loss are both simply integrable in current DL based segmentation approaches and allow to increase the performance for class imbalanced segmentation tasks.
△ Less
Submitted 21 April, 2022; v1 submitted 5 January, 2022;
originally announced January 2022.
-
Sensitivity Analysis for Vehicle Dynamics Models -- An Approach to Model Quality Assessment for Automated Vehicles
Authors:
Marcus Nolte,
Richard Schubert,
Cordula Reisch,
Markus Maurer
Abstract:
Model-based approaches have become increasingly popular in the domain of automated driving. This includes runtime algorithms, such as Model Predictive Control, as well as formal and simulative approaches for the verification of automated vehicle functions. With this trend, the quality of models becomes crucial for automated vehicle safety. Established tools from model theory which can be applied t…
▽ More
Model-based approaches have become increasingly popular in the domain of automated driving. This includes runtime algorithms, such as Model Predictive Control, as well as formal and simulative approaches for the verification of automated vehicle functions. With this trend, the quality of models becomes crucial for automated vehicle safety. Established tools from model theory which can be applied to assure model quality are uncertainty and sensitivity analysis [1]. In this paper, we conduct sensitivity analyses for a single and double track vehicle dynamics model to gain insights about the models' behavior under different operating conditions. We compare the models, point out the most important findings regarding the obtained parameters sensitivities, and provide examples of possible applications of the gained insights.
△ Less
Submitted 11 May, 2020; v1 submitted 8 May, 2020;
originally announced May 2020.
