In Autonomous Driving (AD), accurate positioning is essential to ensure safe, efficient, and cost-effective transportation. However, malicious cyber threats, such as False Data Injection Attacks (FDIA) through Vehicle-To-Everything (V2X) communications, pose significant risks to these objectives. FDIA can manifest as deceptive obfuscation, either translational (T-OBF) – a vehicle falsely claims to be on a parallel road –, or random (R-OBF) – the vehicle reports a random position that differs from its actual position by a few meters. Our approach tackles incorrect position data by leveraging cooperation between trusted participants. These participants perform ML-based FDIA detection by exchanging Angle-Of-Arrival (AoA) measurements, and also for positioning in conjunction with an Extended Kalman Filter to reduce FDIA effects. FDIA detection achieves 89% accuracy for T-OBF and 72% for R-OBF, reducing localization errors below 10m in 77.74% and 81.84% of cases, respectively.

Vehicular platooning aims to improve fuel efficiency and traffic fluidity, and wireless protocols can support such goals. Platoons benefit all vehicle classes (e.g., passenger, truck, trailer), and can be made arbitrarily long, to the extent control and networking mechanisms allow. Vehicle-to-Vehicle (V2V) wireless links inside the platoon are susceptible to path loss, shadowing & diffraction by vehicular obstacles, and fading. In this paper we characterize V2V intra-platoon links at the physical and data-link levels, for platoons composed of same-class vehicles. Our simulation results, using a 30 vehicle-long platoon, show that vehicle dimensions affect propagation and link-level performance, notably that Packet Delivery Ratio (PDR) for edge nodes in an all-passenger vehicle scenario can be 3 times higher than for an all-trailer scenario.

Autonomous driving (AD) is advancing rapidly, but enhancing situational awareness beyond the internal sensors (cameras, radar, and Lidar) requires cooperative perception services enabled by Vehicle-to-Everything (V2X) links. However, some vehicles can inject incorrect information, either intentionally or by hardware malfunction. Wireless positioning offers a set of physical-layer mechanisms, such as Angle-of-Arrival (AoA) and RSS-based ranging, that can verify the correctness of shared data, particularly position data. We introduce Cooperative Wireless Positioning System (CoWiPS), an innovative mechanism in which a set of cooperating vehicles cooperate to estimate the position of a Vehicle-of-Interest. Our positioning solution achieves accurate position estimations, with $70.16 \%$ accuracy within a 10 m error margin under 0dB amplitude noise variance, and $50.01 \%$ accuracy under 5dB variance.

In Autonomous Driving (AD), precise positioning is crucial for safe, efficient and cost-effective transportation. However, malicious cyber threats such as false data injection (FDI) attacks via Vehicle-to-Everything (V2X) links can significantly undermine these goals. FDI involves the insertion of incorrect data into the ego-vehicle or information system (IS), at specific points in time (Impulse FDI) or over specific intervals (Pulse FDI). To combat the specific threat of incorrect position data being shared by a Vehicle-of-Interest, our approach leverages collaboration between a set of trusted participants to perform positioning by exchanging Angle-of-Arrival (AoA) measurements over sidelinks paired with a Kalman Filter. Our method reduces localization errors caused by Impulse FDI to below 10m in 60% of cases, and the errors caused by Pulse FDI in 66.39% of cases.

Wireless positioning (WP) enables ego or hetero-localization, e.g. for asset tracking applications. It can be provided by 3GPP technologies, complementing GPS with extra accuracy where conditions are challenging (e.g., urban canyons). We address scenarios where base-stations (BS) track a mobile user using a wireless Angle-of-Arrival (AoA) technique and, by exchanging angle estimates, can estimate the location of a target node. Cooperative position determination from angle estimates has been addressed by estimation techniques (e.g., Least Squares). We investigate if machine learning techniques, notably a Long Short-Term Memory Neural Network (LSTM-NN), can offer competitive performance. The LSTM is designed to be fed sequential error-affected angle measurements and output position estimates. We assume the path taken by the target mobile node is known beforehand. At training stage, the LSTM is trained to learn a subset of scenario locations (belonging to the node’s path) with well-known positions; then, in runtime, it corrects position estimates for points in the whole path. The LSTM architecture is selected as it retains temporal relationship between input samples. Preliminary simulation results show competitive accuracy when the angle estimate error is relatively large (15°) with respect to a baseline value (5°).

Wireless communications increasingly enable ubiquitous connectivity for a large number of nodes, applications and scenarios. One of the less explored scenarios is aquatic communications, specially when considering near-shore and short-range communications. Overwater communications are impaired by a number of distinguishing dynamic factors, such as tides, waves or node mobility, that lead to a widely fluctuating and unpredictable channel. In this work, we empirically characterize near-shore, overwater channels at 2.4 GHz under realistic conditions, including tidal variations, and relatively short TX-RX separations. To this end, we conducted experiments in a coastal estuarine region and on a harbor to characterize Shore-to-Shore (S2S) and Shore-to-Vessel (S2V) communication channels, respectively, and to identify major factors impairing communication in such scenarios. The empirical results show that constructive/destructive interference patterns, varying reflecting surface, and node mobility (i.e. travel direction and particular maneuvers) have a relevant and noticeable impact on the received signal strength. Thus, a set of parameters should be simultaneously considered for improving the performance of communication systems supporting S2S and S2V links, namely tidal variations, reflection surface changes, antenna height, TX-RX alignment and TX-RX separation. The results useful provide insights into realistic S2S and S2V link design and operation.

This article presents a novel centrality-driven gateway designation framework for the improved real-time performance of low-power wireless sensor networks (WSNs) at system design time. We target time-synchronized channel hopping (TSCH) WSNs with centralized network management and multiple gateways with the objective of enhancing traffic schedulability by design . To this aim, we propose a novel network centrality metric termed minimal-overlap centrality that characterizes the overall number of path overlaps between all the active flows in the network when a given node is selected as gateway. The metric is used as a gateway designation criterion to elect as a gateway the node leading to the minimal number of overlaps. The method is then extended to multiple gateways with the aid of the unsupervised learning method of spectral clustering . Concretely, after a given number of clusters are identified, we use the new metric at each cluster to designate as cluster gateway the node with the least overall number of overlaps. Extensive simulations with random topologies under centralized earliest-deadline-first (EDF) scheduling and shortest-path routing suggest our approach is dominant over traditional centrality metrics from social network analysis, namely, eigenvector , closeness , betweenness , and degree . Notably, our approach reduces by up to 40% the worst-case end-to-end deadline misses achieved by classical centrality-driven gateway designation methods.

Embedded systems (ES) are wide-spread in our world and responsible for many critical systems.More recently, machine learning (ML) tools have become a well-established solution for data-intensive tasks, but their application in embedded systems is still gaining traction and their real-time performance is often unclear.We provide a (non-extensive) review of the ML tools that may be suited for deployment in ES, from which we selected two representative tools -the wellestablished Python-based Scikit-Learn, and the interoperabilityoriented ONNX Runtime -to compare their response time.Using archetypal datasets and four pre-trained ML models, we measure the prediction time for each sample, for each model, in Scikit-Learn and ONNX Runtime in a standard desktop (to compare performance of the tools in the same platform), and for ONNX Runtime in a representative ES, a Raspberry Pi v4 (to compare performance of the same tool across platforms).We report that ONNX considerably improves over Scikit-Learn, and experiences a negligible performance degradation when ported to the RPi.

In the mid-term future, vehicles will generate large amounts of data for both standalone usage (e.g., to recognize road features and external elements such as lanes, signs, and pedestrians) and cooperative usage (e.g., lane merging). However, processing the captured video and image data results comes with significant computational requirements (e.g., GPUs). Computer vision tasks, such as feature extraction, are unfeasible from a business perspective if performed directly in the User Equipment (UE), as automotive manufacturers are unwilling to increase the end-product’s costs. Thus, the logical solution is to collect and upload this data to be processed elsewhere. Nonetheless, processing the data as close to the vehicle is important due to latency constraints, thus calling for the use of Mobile Edge Computing (MEC). An additional benefit of this scenario, in which 5G connectivity enables data to be offloaded to the edge, is that the data from our car is not processed alone. Data from several sources, e.g., multiple vehicles and fixed cameras, can be offloaded to the edge node and processed together, enhancing its quality as more sources of data enhance the prediction output of machine-learning models. This demo showcases a video recording from a vehicle uploaded to an edge node via 5G software-defined-radio FPGA devices. There, a YOLO application to detect objects processes the video and communicates this information to the vehicle, ensuring QoS metrics even when the UE performs handover to a different cell or geographical area.

Intersections are critical elements of urban traffic management and are identified as bottlenecks prone to traffic congestion and accidents. Intelligent intersection management plays a significant role in improving traffic efficiency and safety determining, among other metrics, the waiting time that vehicles incur when crossing an intersection. This work presents a preliminary analysis of the worst-case response time of intersection management protocols that handle mixed traffic with autonomous and human-driven vehicles. We deduce theoretical bounds for such time considered as the interval between the injection of a vehicle in the road system and its departure from the intersection, considering different intersection management protocols for mixed traffic, namely the Synchronous Intersection Management Protocol (SIMP) and several configurations of the conventional Round-Robin (RR) policy. Simulation results validate the analytical bounds partially. Ongoing work addresses the queue dynamics and its reliable detection by traffic simulators.

In the design of shore-to-shore and shore-to-vessel wireless links, the impact of the ray reflected on the surface is often neglected. It adds that, in some coastal areas, the geometry of the reflection changes over time due to tides. When choosing an antenna height for an inshore node, often the largest possible height is used, but this approach can lead to signal degradation. The two-ray model is the most fundamental path loss model to account for the contribution of the reflected ray. We carried out experimental measurements at the shores of a freshwater body to verify that the two-ray model can predict the major trends of the path loss experienced by a 2.4 GHz over-water wireless link. We focus on short-to-medium distance links, with antennas installed a few meters above surface. We observed considerable consistency between measurements and model estimates, leading us to conclude that the two-ray model may bring benefits when applied to the network design of over-water links affected by tidal variations, which is our end-goal.

The rising adoption of wireless technologies in the In- dustrial Internet of Things has stressed the need for traffic schedulability validation at system design-time to support safety and time critical streams (e.g., process control and emergency response). In this context, the demand-based schedulability tests have recently been proposed in the literature. This work revisits two well-established techniques borrowed from the multi-processor scheduling theory, namely the demand-bound-function (DBF) and the forced-forward-demand-bound-function (FFDBF), and evaluates their performances when adapted to the field of wireless sensor-actuator networks. Simulation experiments when varying network configurations confirm the equal or better accuracy of FFDBF over DBF to estimate both network demand and schedulability. In future work, we aim at building upon these promising results in order to design novel admission control and adaptation strategies that improve network schedulability under varying workload conditions.

In-vehicle wireless networks (e.g., Wi-Fi, Bluetooth) are experiencing a faster market penetration than dedicated V2V technologies and are compatible with non-V2X devices. In this paper, we assess whether commodity in-vehicle networks can leverage opportunistic V2X communication to nodes outside of the vehicle, particularly with Vulnerable Road Users. We characterize the radiation pattern and performance of communication links in the 2.4 GHz band between in-car wireless networks and a wireless-enabled bicycle in two representative interaction scenarios (i.e. parallel and perpendicular vehicle-VRU trajectories) using both production hardware (i.e. built-in WiFi hotspot) and dedicated measurement equipment. Empirical results show that (i) the signal propagation to the outside of the vehicle is strongly affected (up to 20 dB) by the vehicle elements (e.g. pillars) and by the placement of the wireless system inside the car, and (ii) the communication performance (in terms of RSSI, IRT, Throughput) is also impaired by the spatial arrangement of vehicle and VRU, and other time-varying phenomena (e.g., human body and bicycle shadowing). We conclude that the in-car system performance allows supporting a wide range of safety and infotainment applications (e.g., IRT under 300 ms) even at large TX-RX distances, and that the placement of an in-car wireless system should be tailored according to the target application.

Bicycle-to-bicycle (Bi2Bi) communication can be implemented by well-established technologies in the 2.4GHz ISM band: IEEE 802.11, Bluetooth or IEEE 802.15.4. These technologies have distinct performance due to different physical and data link layers. In this paper, we characterize the mentioned 2.4 GHz-operating technologies over opportunistic links established between bicycles using commodity hardware. We find that, in Bi2Bi links, Blue-tooth, IEEE 802.11 at 24 Mbit/s, and IEEE 802.11 with automatic rate adaptation can communicate only in the immediate surroundings (under 15m of range), to maxima of 1.5 Mbit/s, 17 Mbit/s and 25 Mbit/s, respectively. IEEE 802.15.4 and IEEE 802.11 at 1 Mbit/s sustain connectivity up to 30 and 40 meters and peak transfer rates of 50 kbit/s and 800 kbit/s respectively. In addition, we observed that, in all measurement scenarios, link performance depended strongly on whether bicycles were approaching or moving away, rather than on whether one was at the front or back of the other.

Smart cities aim to improve the citizens' quality of life by leveraging information about urban scale processes extracted from heterogeneous data sources collected on citywide deployments. The Internet-of-Things (IoT) is, thus, the enabler of smart city technologies at urban scale. In this paper, we present PortoLivingLab, a multisource sensing infrastructure that leverages IoT technology to achieve city-scale sensing of four phenomena: weather, environment, public transport, and people flows. To sense these processes on a city scale, we deployed a vehicular network with over 600 vehicles and 19 static environmental sensors. We also developed an easily reconfigurable crowdsensing platform and carried out several crowdsensing campaigns with more than 600 participants. The data is collected in a common backend and stored using similar spatio-temporal data models to simplify sharing and joint analysis for the characterization of urban dynamics. We describe the architecture and composing elements of PortoLivingLab, highlighting the IoT technologies, and challenges faced. We present several proof-of-concept use cases (e.g., passenger flows from WiFi connections) that provide new insights into different components of an evolving and moving city. Finally, we lay out the future lines of work that will strive for finding hidden phenomena by leveraging data from the three complementary platforms.

A large-scale urban sensing platform, composed of multiple Data Collection Units (DCUs) equipped with sensor hardware scattered accross the city, allows pervasive monitoring of environmental parameters. Gathering sensor data from a number of disparate locations at a backend server can be supported by delay-tolerant services provided by existing vehicular networks. Our real-world sensing platform takes advantage of an existing vehicular network with more than 400 vehicles equipped with On-Board Units (OBUs). A purposely-developed implementation of a delay tolerant service is installed in all elements involved in the communication flow, from DCUs to the backend server. In this demo, we showcase the full end-to-end data flow with the actual equipment being used in our real-world deployment. Data produced at a DCU is collected by an OBU installed in a vehicle and delivered to a Road-Side Unit (RSU), which then forwards the data to the backend server.

Many wireless applications require a propagation model that describes the attenuation of the transmitted signal as a function of the distance between devices. Such channel models are derived commonly from signal strength measurements, and assume that the true distances between wireless terminals are known. In practice, however, the true distances may be unavailable or difficult to obtain, for instance in mobile scenarios or in the absence of line-of-sight. These conditions typically occur in forested environments, urban areas, etc. This paper addresses the problem of path loss model parameter estimation in presence of erroneous distance measurements, such as the ones derived from the GPS positions. We provide a model for the uncertainties, and study the impact of distance errors on the estimation of a log-distance channel model. Our main conclusion is that the path loss model can be estimated with a reasonable accuracy from unreliable distances, provided that the measurements are taken at distances beyond a few standard deviations of the GPS positioning error. In case the maximum communication range does not allow such large distances, we provide a method to correct the erroneous channel model. Real-world measurements are used in order to validate our approach.