TTTech Computertechnik (Austria)

With the rapid development and integration of the Internet of Things into every aspect of daily life, finding more intuitive and natural methods of human-computer interaction has become increasingly important. Hand gesture detection-based systems are garnering significant attention in this field, as they offer a natural and simpler way of communicating with various devices compared to traditional input methods like keyboards and mice. One approach to detecting user hand gestures involves generating a non-audible sound through speakers and analyzing the reflected signal using microphones. This paper proposes a solution for detecting 11 distinct hand gestures based on deep learning and the Doppler effect of sound waves. The system utilizes commercial speakers and microphones to generate a 20 kHz audio signal and analyze its reflection. The detection model is a Convolutional Neural Network (CNN), trained on a dataset of 2100 samples and tested on an additional 550 samples. The model achieves 91.43% accuracy and demonstrates high performance when evaluated on previously established datasets.

Building Information Modeling (BIM) has been widely adopted for construction inspections due to its ability to integrate multiple data sources. Engineers use BIM to identify and review site issues, yet inspection systems face several challenges. Firstly, positioning inspection areas on a construction site using BIM with Augmented Reality (AR) requires complex model manipulation. Additionally, signal or Internet connectivity issues may limit positioning technologies. Secondly, human error or interference is common in traditional inspection processes due to their complexity. To overcome these barriers, this research applied BIM and AR with Visual Simultaneous Localization and Mapping (VSLAM) to help inspectors quickly and effectively record construction defects as photographs with notes and their locations. An efficient approach is proposed to integrate BIM and AR with VSLAM, and a prototype is developed to validate and demonstrate how the proposed system can assist a site inspector in performing quality management, even offline. The system uses a two-phase indoor positioning method: initial localization via visual markers and real-time tracking with VSLAM, enabling precise defect tracking and efficient model adjustments. While significantly improving inspection accuracy and efficiency, its performance is affected by environmental factors like lighting and marker placement, providing insights for future refinement.

This paper presents a field programmable gate array (FPGA)–based implementation of a deep neural network (DNN) for handwritten digit recognition. We propose the use of a fully connected four‐layer neural network with the hidden layers implementing the ReLU activation function and the output layer based on the Softmax activation function. The neural network model, including the forward propagation algorithm and the backward propagation (BP) algorithm, is entirely implemented in an FPGA system, enabling both testing and training of the network. Mathematical operations in the network are performed on 32‐bit floating‐point data. The proposed network achieves promising results in terms of system simplicity and low consumption of hardware resources while maintaining acceptable recognition accuracy and algorithm execution speed at the same time. The system is implemented on Xilinx’s ZYBO development board and achieves a precision of 92.8% digit recognition. Training and test images for system evaluation are obtained from the MNIST database. The system occupies 10,020 (57%) lookup tables, 7781 (22%) registers, 17 (21%) DSP blocks, and 3 (5%) BRAM blocks. The maximum power consumption of the system is equal to 1.602 W, at the operating frequency of the system equal to 100 MHz.

The increasing use of acceleration hardware, such as DSPs, FPGAs and GPUs, has significantly boosted computational performance potential in various domains. However, in safety-critical systems, this hardware can pose risks, especially in mixed-criticality scenarios where non-critical and non-preemptible workloads can delay the execution of critical tasks. To address this challenge, we introduce a vendor-independent Hardware Abstraction Layer (HAL) framework based on Vulkan that enables preemptive execution handling of compute shaders for devices for which a Vulkan driver is available. This mechanism shows potential for timely execution of critical workloads, making it a valuable tool for embedded systems, low-power designs, and machine learning. Our prototype framework utilizes modified GLSL shader source code to support preemption, allowing shaders to exit before completing all work steps if a stop-flag is set by the host system. While Vulkan support is not restricted to GPUs we focus our work on these devices and validate our framework through benchmarks on an NVIDIA GeForce GTX 1080 Ti GPU, comparing performance across various shader types. Results show that our preemptive shaders, though slightly slower due frequent polling of the stop-flag, offer a promising solution for integrating acceleration hardware into safety-critical and real-time heterogeneous compute systems.

The automotive sector is undergoing profound changes with the advancement of technology, and connected vehicles represent one of the most notable examples. These vehicles, equipped with internet connectivity and communication capabilities with other devices, are becoming increasingly widespread. Consequently, they accumulate substantial amounts of data concerning drivers and their environments. However, this connectivity also brings about significant security concerns, particularly regarding the privacy and security of the metadata stored in these vehicles. Metadata encompasses diverse information about user activities, habits, location data, and personal preferences, making it an appealing target for potential attackers. Therefore, safeguarding the security of metadata in connected vehicles stands out as a primary concern for manufacturers. This study aims to surpass the two-factor authentication (2FA) systems developed to protect the metadata stored in connected vehicles. The system comprises two components: the Central Security Unit (CSU) and the AutoGuard (AG) mobile application. Integrated with the Remote Keyless Entry System (RKES), CSU initiates the 2FA process when the driver approaches the vehicle. Upon entering the second authentication factor (biometric, pattern, PIN code), successful authentication unlocks the vehicle doors via RKES, while failure prompts notification to the driver. To advance the 2FA system, a Bluetooth Low Energy (BLE)-based communication system has been integrated between CSU and AG. This integration enhances communication between the two, making it more secure and energy efficient. Furthermore, it enables data exchange without the need for a network connection, ensuring a seamless user experience. These innovative communication features transcend the 2FA system by rendering communication between CSU and AG more reliable and flexible. Additionally, by integrating the phone's location into the authentication system, AG functionality can enhance the accuracy of the 2FA system, potentially using the phone's location as an additional authentication factor.

We introduce a new paradigm for synthesizing time-triggered schedules that guarantees the correct temporal behavior of time-triggered (TT) tasks and the schedulability of sporadic event-triggered (ET) tasks with arbitrary deadlines at design time. The approach first expresses a constraint for the TT task schedule in the form of a maximal affine envelope that ensures that as long as the schedule generation respects this envelope, all sporadic ET tasks meet their deadline. The second step consists of modeling this envelope as a burst limiting constraint (BLC) and building the schedule. The BLC constraint can be added to any existing TT schedule generation method as an additional constraint on the TT slot positioning. Here, we propose an efficient TT schedule generation method that integrates the BLC constraint via simulating a modified Least-Laxity-First (LLF) scheduler. We show via synthetic and real-world test cases that our novel method achieves better schedulability and a faster schedule generation for most use cases compared to other approaches inspired by, e.g., hierarchical scheduling. Moreover, we present an extension to our method that finds the most favorable schedule for TT tasks with respect to ET schedulability, thus increasing the probability that the system remains feasible when ET tasks are later added or changed.

To encourage the uptake of AI in industrial use cases, tools are required to support the engineering process throughout the life cycle of the AI application that is central to the use cases. Providing guidance on using AI in industrial setups is vital for creating trustworthy, reliable, and ethically compliant AI-based solutions. The related standardization landscape and available guideline repositories are large, scattered over the web, change rapidly, and are hard to keep up with. This limits the access and ease of use of these standards and guidelines. To address these limitations, we propose developing a self-assessment tool, empowered through AI algorithms and models such as Large Language Models, to improve the process of accessing and benefiting from those standards and guidelines. This self-assessment tool will support various user groups while engineering their applications by identifying the most applicable guidelines according to the individual attributes of the specific user group. We argue that modeling specific attributes and mapping appropriate controls for self-assessments could be achieved by applying AI-based technologies. This paper outlines our ongoing efforts concerning the suggested supporting tools, offering a human-centric methodology. Additionally, we present initial results demonstrating how the needs of a particular user group can be accurately modeled. The results of this study will be used for applications that are deploying AI in an industrial setting with the objective of enabling the two most important goals of Industry 5.0, which are the well-being of workers at the center of the production process, and sustainable and resilient industries.

Container-based virtualization is a promising deployment model in fog and edge computing applications, because it allows a seamless co-existence of virtualized applications in a heterogeneous environment without introducing significant overhead. Certain application domains (e.g., industrial automation, automotive, or aerospace) mandate that applications exhibit a certain degree of temporal predictability. Container-based virtualization cannot be easily used for such applications, since the technology is not designed to support real-time properties and handle temporal disturbances. This article proposes a framework consisting of a static offline and a dynamic online phase for resource allocation and adaptive re-dimensioning of real-time containers. In the offline phase, the optimal initial deployment and dimensioning of containers are decided based on ideal system models. Additionally, to adapt to dynamic variations caused by changing workloads or interferences, the online phase adapts the CPU usage and limits of real-time containers at runtime to improve the real-time behavior of the real-time containerized applications while optimizing resource usage. We implement the framework in a real Linux-based system and show through a series of experiments that the proposed framework is able to adjust and re-distribute computing resources between containers to improve the real-time behavior of containerized applications in the presence of temporal disturbances while optimizing resource usage.