This paper evaluates the current maturity of automatic code-generation workflows for deploying modern CNN-based object detectors on embedded GPU platforms. We compare a native pipeline against a code generation pipeline through a Model-Based Engineering (MBE) approach, using YOLOv8/YOLOv9 inference on NVIDIA Jetson Orin Nano and Jetson AGX Orin as representative edge-GPU workloads. We report detection-quality metrics (mAP, PR curves) and system-level metrics (latency distribution and initialization overhead) under a controlled single-class scenario based on a CARLA-generated sequence with frame-level annotations. Absolute accuracy and latency values are scenario-dependent and may vary under different camera optics, illumination, motion blur, sensor noise, occlusion patterns, and multi-class scene. Results quantify the performance gap between code generation and native pipelines and show that, for the evaluated workloads, the automated pipeline remains less competitive in both latency and accuracy. We discuss the implications of this gap for deployment workflows in safety-oriented domains, and we outline bottlenecks that should be addressed. The study is intended as a controlled traffic-light detection micro-benchmark and does not aim to validate full ADAS perception stacks.
Transforming decision-making in the certification of aerostructures through AI: Inside the European pAIramid project pAIramid provides AI‑enabled virtual tools to support decision‑making in aerostructure certification, optimising design and manufacturing strategies with reduced physical testing. The certification of composite aerostructures remains one of the most time-consuming and costly stages in the development of new aircraft. Traditional certification relies on a sequential testing pyramid – material, coupon, element, and full aerostructure – requiring extensive physical testing and repeated validation. This rigid methodology slows down innovation and limits the ability to introduce new materials or manufacturing processes. The European project pAIramid aims to address these challenges by creating a decision-making tool that will help users to define designs and process guidelines from any level of the certification test pyramid.
Accurate 6D pose estimation of complex objects in 3D environments is essential for effective robotic manipulation. Yet, existing benchmarks fall short in evaluating 6D pose estimation methods under realistic industrial conditions, as most datasets focus on household objects in domestic settings, while the few available industrial datasets are limited to artificial setups with objects placed on tables. To bridge this gap, we introduce CHIP, the first dataset designed for 6D pose estimation of chairs manipulated by a robotic arm in a real-world industrial environment. CHIP includes seven distinct chairs captured using three different RGBD sensing technologies and presents unique challenges, such as distractor objects with fine-grained differences and severe occlusions caused by the robotic arm and human operators. CHIP comprises 77,811 RGBD images annotated with ground-truth 6D poses automatically derived from the robot's kinematics, averaging 11,115 annotations per chair. We benchmark CHIP using three zero-shot 6D pose estimation methods, assessing performance across different sensor types, localization priors, and occlusion levels. Results show substantial room for improvement, highlighting the unique challenges posed by the dataset. CHIP will be publicly released.
Induction hardening is a heat treatment that has been increasingly employed in the industry in recent years. It is a complex, highly coupled, and multiphysical process involving electromagnetism, thermal, mechanical, and metallurgical physics. One of the main quality requirements of the process is the hardened case depth generated in the workpiece. The usual method to measure the hardened case and ensure the quality of the parts is to use destructive techniques, which generate material and energy waste and production inefficiencies. Additionally, selecting process parameters such as current, frequency, or scanning speed typically requires several trial-and-error iterations. The goal of this work is to provide a hybrid digital twin (DT) that acts as a nondestructive test technique, predicting the resulting hardened case in real-time and enabling the correction of process parameters during the induction hardening process, ultimately achieving a zero-waste manufacturing scheme. For this purpose, a DT based on an artificial neural network (ANN) model is developed, predicting the hardened case depth in real-time using four monitored input variables: induction frequency, current, and two temperature measurements on the surface of the hardened part. The required data for DT development and training are obtained using a finite element model. Several ANN architectures are evaluated, and the configuration with the best regression results is chosen for implementation in an industrial induction hardening machine. The hardened case predictions obtained from the developed DT demonstrate high accuracy within the analyzed frequency and current range.
Fuzzing, a powerful tool for vulnerability discovery, is limited by the coarse-grained, binary nature of its crash detection oracle. The use of sanitizers strengthens this oracle but requires recompilation or binary rewriting, and is limited to known patterns of vulnerabilities. We investigate an alternative way to strengthen the implicit oracle that is suitable for small (IoT-sized) devices: electromagnetic (EM) side-channel analysis. By integrating this into a fuzzing campaign we are able to detect anomalous program states through physical execution patterns. GAFLERNA, our EM-enhanced AFL++ framework, achieves a 87% correlation with sanitizer findings in the best case, without modifying the executable, while discovering 104 new paths to known crashes across four real-world programs. This reveals the potential for hardware-level feedback to extend fuzzing and analyse IoT programs where only the binary code is available.
