The automated generation of diversified training scenarios has been an important ingredient in many complex learning tasks, especially in real-world application domains such as autonomous driving, where auto-curriculum generation is considered vital for obtaining robust and general policies. However, crafting traffic scenarios with multiple, heterogeneous agents is typically considered a tedious and time-consuming task, especially in more complex simulation environments. To this end, we introduce MATS-Gym, a multi-agent training framework for autonomous driving that uses partial-scenario specifications to generate traffic scenarios with a variable number of agents which are executed in CARLA, a high-fidelity driving simulator. MATS-Gym reconciles scenario execution engines, such as Scenic and ScenarioRunner, with established multi-agent training frameworks where the interaction between the environment and the agents is modeled as a partially observable stochastic game. Furthermore, we integrate MATSGym with techniques from unsupervised environment design to automate the generation of adaptive auto-curricula, which is the first application of such algorithms to the domain of autonomous driving. The code is available at https://github.com/AutonomousDrivingExaminer/mats-gym.

Understanding attention is crucial for improving safety in driving scenarios. Detected and classified objects, along with their observation by the driver, are used as a measure of attention. This paper investigates the differences between human and artificial attention in real-world and replay driving scenarios. By analyzing attention patterns from drivers and a vision-language model agent, we identify a number of differences. The results highlight the limitations of current AI attention models and suggest the way forward for developing more context-aware systems.

Video moment retrieval aims to locate the timestamps best matching the query description within an untrimmed video. However, existing video moment retrieval approaches typically suffer from two major limitations: (1)Utilize only negative moment-sentence pairs sampled from intra-videos, which may overfit the bias of the dataset and not have an excellent understanding of the video and query due to the dataset size and annotation biases. (2)Decouple the video and the query, perform unimodal learning separately, and then concatenate them together as multimodal fusion features. In this paper, we propose a novel approach named Momentum Contrastive Matching Network(MCMN). Inspired by MoCo, we propose the Momentum Cross-modal Contrast for cross-modal learning to enable large-scale negative sample interactions, which contributes to the generation of more precise and discriminative representations, and use temporal decay to model key attenuation in the memory queue when computing the contrastive loss. In addition, we use an attention module to adaptively generate clip-specific word embeddings to achieve semantic alignment from a temporal perspective, which are considered to be more important for finding relevant video contents with large boundary ambiguities. Experimental results on the three major video moment retrieval benchmark datasets, including TACoS, Charades-STA, and ActivityNet Captions demonstrate that MCMN surpasses previous methods and reaches state-of-the-art with disparate visual features.

The NimbleAI Horizon Europe project leverages key principles of energy-efficient visual sensing and processing in biological eyes and brains, and harnesses the latest advances in$\mathbf{33D}$stacked silicon integration, to create an integral sensing-processing neuromorphic architecture that efficiently and accurately runs computer vision algorithms in area-constrained endpoint chips. The rationale behind the NimbleAI architecture is: sense data only with high information value and discard data as soon as they are found not to be useful for the application (in a given context). The NimbleAI sensing-processing architecture is to be specialized after-deployment by tunning system-level trade-offs for each particular computer vision algorithm and deployment environment. The objectives of NimbleAI are: (1)$\mathbf{100x}$performance per mW gains compared to state-of-the-practice solutions (i.e., CPU/GPUs processing frame-based video); (2)$\mathbf{50x}$processing latency reduction compared to CPU/GPUs; (3) energy consumption in the order of tens of mWs; and (4) silicon area of approx. 50 mm2.

In this paper we present a hardware-in-the-loop (HiL) framework for testing wireless vehicle-to-everything (V2X) communication hardware, i.e., modems under realistic channel conditions. The framework includes a wireless channel emulator, which is capable of emulating non-stationary wireless channels in real-time. We validate the HiL framework by comparing the frame error rate (FER) obtained via emulation with data obtained during a V2X measurement campaign using the same IEEE 802.11p based modems. To do this we acquire measured time-variant channel transfer function and FER measurements simultaneously. The results show that our HiL approach is feasible and that we can obtain FER measurements in the laboratory that closely match the measurement results obtained on the road, giving the maximal distance of 0.099 between their cumulative distribution functions.

This paper discusses a freely available and open dataset containing vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I) and vehicle-to-pedestrian (V2P) OFDM-based wireless channel measurement data including synchronised sensor data such as radar, LiDAR and high precision GPS. The wireless channel measurement is conducted at the carrier frequencies of 3.2 GHz and 5.81 GHz which are the most promising frequency bands in which future V2X communication systems will operate. The dataset contains the wireless channel measurement data of various V2X scenarios along with synchronized sensor information from a vehicle. In addition to the wireless channel measurement data, the dataset also includes frame error rate measurements from a IEEE 802.11p based communication system, synchronized to the other measurement data.

In this paper we adopt and modify a well-known locality-aware hashing scheme to the problem of active stochastic scatterer selection in vehicular non-stationary geometry-based stochastic channel models (GSCM). We show, how under relaxed assumptions on the query set an efficient selection of active stochastic scatterers during simulation is computationally feasible. The proposed approach enables real-time simulation and emulation of large-scale GSCMs by restricting the active stochastic scatterer set to meet given resource constraints. We showcase our approach by introducing a GSCM that is boot-strapped via OpenStreetMap data. The stochastic scatterers are placed automatically along buildings, traffic signs and vegetation. We validate and investigate the impact of the proposed approach on the accuracy of a GSCM by means of second order statistics of the time- and frequency-varying fading process. For validation and performance evaluation we parameterize our GSCM using a vehicular wireless channel measurement campaign conducted in the inner city of Vienna. The impact of selecting only a subset of scatterers is then evaluated using the calibrated GSCM.

Wireless embedded devices are key elements of Internet-of-Things (IoT) and industrial IoT (IIoT) applications. The complexity of these devices as well as the number of connected devices to networks increase steadily. The high intricacy of the overall system makes it error-prone and vulnerable to attacks and leads to the need to test individual parts or even the whole system. Therefore, this paper presents the concept of a flexible and distributed testbed to evaluate correct behavior in various operation or attack scenarios. It is based on the Robot Operating System (ROS) as communication framework to ensure modularity and expandability. The testbed integrates RF-jamming and measurement devices to evaluate remote attack scenarios and interference issues. An energy harvesting emulation cell is used to evaluate different real-world energy harvesting scenarios. A climatic test chamber allows to investigate the influence of temperature and humidity conditions on the system-under-test. As a testbed application scenario, the automated evaluation of an energy harvesting wireless sensor network designed to instrument automotive engine test benches is presented.

Electronic data-sheets that are stored within and provided by a transducer such as ISO/IEC/IEEE 21450 Transducer Electronic Data-sheets are currently intended to ease the utilization of transducers. This is achieved as the data-sheets provide information about sensing and actuation as well as communication capabilities, data formats, calibration information and more. However, so far it can not be ensured that the provided information is trustworthy. Consequently, we propose a concept based on ISO/IEC/IEEE 21450 Transducer Electronic Data-sheets in which signatures and secure elements are used to validate the information in the data-sheets, the manufacturer and calibration labs. Furthermore, this information can be utilised for secure key exchange mechanisms and permission management. This approach ensures that the low-power consumption expected of transducers is still met while providing security and trust.

Secure wireless communication is essential for most industrial applications. The secure and reliable control of processes as well as the data integrity of measured values are key targets in these applications. The industrial Internet-of-Things (IIoT) tries to connect an increasing number of sensors wirelessly. The wireless sensors form wireless sensor networks (WSNs). However, wireless sensor nodes are exposed to various security threats ranging from physical modification on the device itself to remote attacks via the communication channel. It is important to secure the complete lifetime of the wireless sensor nodes and other system components. This includes the production phase, shipping, preparation phase and operational phase. This paper presents a lifetime security concept for a wireless sensor network applied in automotive test beds. In this application scenario, the wireless sensor nodes are used to capture various temperatures in an automotive unit under test. In order to indicate the current state of trustworthiness of the system, a trustworthiness indicator is implemented which is shown to the user. An evaluation of the impact of encrypted communication on power consumption shows that the increase is negligible, and can be expected to be provided by the wireless sensor node's power supply without reducing the node lifetime.

For distributed measurements to be consistently aligned a global notion of time within the network is necessary. Clock synchronization performance is hampered by delays and jitter accumulated not only in the network, but also in the timestamping procedures of the devices being synchronized. The resources available on such sensors are often constrained, both by cost and power consumption. Software solutions on single processor systems are common. While the synchronization process is critical, it usually has a supporting role and therefore cannot be assumed to have the highest priority. This paper analyzes the influence of periodic, non-interruptable tasks (e.g. measurements) on the timestamping process, presents mathematical models for simulation and a modified median filter to efficiently deal with these errors.

Wireless sensor networks (WSNs) currently see a strong increase in demand for various industrial applications like smart testing and production systems. Open frequency bands (e.g. 2.4 GHz) allow the use of commercial low-cost hardware, but need to be shared with other devices and systems, leading to interference and therefore transmission errors. Re-transmitting damaged packages might eventually succeed, but at the costs of increased latency and additional energy consumption of the node. As minimizing energy consumption is paramount for self-owered WSN nodes, interference avoidance is highly desired. In this work a channel management procedure for WSNs is proposed which can react to dynamically changing interference scenarios. A short monitoring period is used to sample received signal strength indication values of a set of candidate frequency bands. An extensive measurement campaign is performed and six analysis methods are compared, and effects of applying the resulting choice of “best” frequency bands to WSNs in an automotive testing environment are investigated. The best method allows a reduction of energy consumption of a node by about 15%.

The ARTEMIS 1 project DEWI ("Dependable Embedded Wireless Infrastructure") focusses on the area of wireless sensor / actuator networks and wireless communication. With its four industrial domains (Aeronautics, Automotive, Rail, and Building) and 21 clearly industry-driven use cases / applications, DEWI will provide and demonstrate key solutions for wireless seamless connectivity and interoperability in smart cities and infrastructures, by considering everyday physical environments of citizens in buildings, cars, trains and airplanes. It will add clear cross-domain benefits in terms of re-usability of techno-logical building bricks and architecture, processes and methods. DEWI currently is one of the largest funded European R&D projects, comprising 58 renowned industrial and research partners from 11 European countries. (For further details see www.dewi-project.eu)

In 2011 Ericsson and Cisco estimated 50 billion Internet connected devices by 2020, encouraged by this industry is developing application frameworks to scale the Internet of Things. This paper presents a survey of commercial frameworks and platforms designed for developing and running Internet of Things applications. The survey covers frameworks supported by big players in the software and electronics industries. The frameworks are evaluated against criteria such as architectural approach, industry support, standards based protocols and interoperability, security, hardware requirements, governance and support for rapid application development. There is a multitude of frameworks available and here a total 17 frameworks and platforms are considered. The intention of this paper is to present recent developments in commercial IoT frameworks and furthermore, identify trends in the current design of frameworks for the Internet of Things; enabling massively connected cyber physical systems.

Increasing the efficiency of production and manufacturing processes is a key goal of initiatives like Industry 4.0. Within the context of the European research project ARROWHEAD, we enable and secure smart maintenance services. An overall goal is to proactively predict and optimize the Maintenance, Repair and Operations (MRO) processes carried out by a device maintainer, for industrial devices deployed at the customer. Therefore it is necessary to centrally acquire maintenance relevant equipment status data from remotely located devices over the Internet. Consequently, security and privacy issues arise from connecting devices to the Internet, and sending data from customer sites to the maintainer's back-end. In this paper we consider an exemplary automotive use case with an AVL Particle Counter (APC) as device. The APC transmits its status information by means of a fingerprint via the publish-subscribe protocol Message Queue Telemetry Transport (MQTT) to an MQTT Information Broker in the remotely located AVL back-end. In a threat analysis we focus on the MQTT routing information asset and identify two elementary security goals in regard to client authentication. Consequently we propose a system architecture incorporating a hardware security controller that processes the Transport Layer Security (TLS) client authentication step. We validate the feasibility of the concept by means of a prototype implementation. Experimental results indicate that no significant performance impact is imposed by the hardware security element. The security evaluation confirms the advanced security of our system, which we believe lays the foundation for security and privacy in future smart service infrastructures.

Europa has launched multiple initiatives and research projects to remain competitive in a globalized world and keep industry and manufacturing on-shore. Funded by EU and member countries, project ARROWHEAD[1] focuses research and innovation for collaborative automation using interoperable services for smart production, to improve quality, efficiency, flexibility and cost competiveness. This includes an important new aspect called “Smart Services”, which aims to apply SOA (service oriented architecture) to maintenance and service of production systems and its parts, which still carry a huge potential for further gains in cost and energy savings. However, there will be no “big bang”. How can we turn present-day variety of diverse, specialized, and legacy loaded embedded systems into connected, SOA based cooperating participants of the Internet of Things (IoT)? This case study portrays the solution followed in ARROWHEAD WP1.1, for devices used in end-of-line (EoL) test systems in automotive powertrain production.