Achieving real-time performance is an important goal for automated driving and ADAS applications. One optimisation for such systems is the use of lower resolution images for CNN based object detection, which can greatly improve inference speed. Reducing image resolution reduces the size of the image but also reduces image quality. The downsampling method used in ADAS is a topic often not considered when downsizing images, and this study aims to address this gap. This study investigates how downsampling using different interpolation methods impacts machine vision performance. Several common machine vision algorithms are trained on downsampled data, and their performance is evaluated. The downsampling methods used are: Bilinear interpolation, Bicubic interpolation, Nearest Neighbour interpolation, Area-Based resampling and Lanczos4 interpolation. The results show that training with different downsampling methods does have a consistent impact on performance across different object detection algorithms; however, the differences are generally very small, with a difference of less than 2% AP50 in most cases. One object detection model (RT-DETR) is shown to be much more sensitive to interpolation methods. This study indicates which methods of downsampling are best suited for use in ADAS applications, and their relative advantages and disadvantages of each method. The results presented here are relevant to designers of ADAS who are concerned with real-time optimisations.
Repeated driving along the same route is common in the real world and promotes familiarity, which significantly influences driving behavior and safety. While prior studies have characterized familiarity using aggregated route-level trends, little attention has been given to how it manifests across diverse scenarios and individuals. This study adopts a counterexample-driven approach to examine whether such global summaries adequately reflect nuanced behavioral adaptations across contexts and drivers. A preliminary study was conducted using a driving simulator with eye tracking. Participants repeatedly drove a virtual urban route containing intersections, curves, and regulatory signs. Gaze behavior was analyzed using stationary and dynamic metrics to capture changes in visual attention. Results show that while global route-level trends indicate increased driving speed and decreased gaze entropy, scenario- and individual-level analyses reveal substantial variability. Notably, the common assumption that fixation duration on traffic signs uniformly decreases with increasing familiarity is challenged, as one subject continued using stop signs as visual references for stopping. These findings highlight the limitations of relying solely on global analyses and underscore the value of incorporating scenario- and individual-level perspectives to better understand how driving familiarity develops. This shift may inform the design of adaptive, human-centric autonomous systems with improved safety and efficiency.
For the verification and validation of Automated Driving System features, vehicle manufacturers use scenario-based approaches to break down the world into more manageable traffic scenarios. Since this approach is typically independent from systems engineering processes such as requirements engineering, the Scenario as Specification approach was proposed, which unifies both approaches by using scenarios as a central development artifact. This work focuses on the integration of a technology-independent and technology-dependent safety engineering to the Scenario as Specification approach.Therefore, we present (1) an extension of the Scenario as Specification approach by introducing a technology-independent feature-level and a technology-dependent system-level view, (2) a deeper integration of Functional Safety and Safety Of The Intended Functionality, (3) the corresponding adaptations of the scenario structure, and (4) an end-to-end practical example with a focus on safety.Our work aims to lay the foundation for a unified scenario-based approach in industry and research communities for the development and deployment of Automated Driving System features.
This paper presents the development of a fuzzy logic-based dual adaptive model predictive control (AMPC) framework for robust autonomous vehicle (AV) path tracking under varying road conditions. The framework integrates two AMPCs: one for lateral-yaw control and another for longitudinal control. The lateral controller regulates AV position and yaw, while the longitudinal controller manages the vehicle speed. A fuzzy inference system dynamically adjusts the lateral controller weights, and the reference longitudinal velocity based on specific road condition inputs, including road friction and path curvature. The controllers were designed and implemented in MATLAB and Simulink and evaluated using a standard double-lane-change manoeuvre. The results demonstrate the effectiveness of the proposed framework in achieving integrated trajectory tracking and stability control for AVs.
This paper presents the design and implementation of a robust and modular teleoperation system for Intelligent Vehicles (IVs), intended to serve as a fallback solution when autonomous vehicles (AVs) encounter operational limitations. Built using modern technologies such as WebRTC, LTE 4G and node-based architectures, the system enables real-time remote driving with low latency and high reliability. The proposed solution was deployed and validated on a real research vehicle, controlled from a remote teleoperation station. Both qualitative and quantitative evaluations were performed, including latency and usability analysis. The results demonstrate that the system is reliable, scalable and suitable for real-world teleoperation applications, offering a solid foundation for future research in AVs and remote driving systems.
