In this paper, we present Vi-Fi, a multi-modal system that leverages a user's smartphone WiFi Fine Timing Measurements (FTM) and inertial measurement unit (IMU) sensor data to associate the user detected on a camera footage with their corresponding smartphone identifier (e.g. WiFi MAC address). Our approach uses a recurrent multi-modal deep neural network that exploits FTM and IMU measurements along with distance between user and camera (depth information) to learn affinity matrices. As a baseline method for comparison, we also present a traditional non deep learning approach that uses bipartite graph matching. To facilitate evaluation, we collected a multi-modal dataset that comprises camera videos with depth information (RGB-D), WiFi FTM and IMU measurements for multiple participants at diverse real-world settings. Using association accuracy as the key metric for evaluating the fidelity of Vi-Fi in associating human users on camera feed with their phone IDs, we show that Vi-Fi achieves between 81% (real-time) to 91% (offline) association accuracy.

We compare the potential of neural network (NN)-based channel estimation with classical linear minimum mean square error (LMMSE)-based estimators, also known as Wiener filtering. For this, we propose a low-complexity recurrent neural network (RNN)-based estimator that allows channel equalization of a sequence of channel observations based on independent time- and frequency-domain long short-term memory (LSTM) cells. Motivated by Vehicle-to-Everything (V2X) applications, we simulate time- and frequency-selective channels with orthogonal frequency division multiplex (OFDM) and extend our channel models in such a way that a continuous degradation from line-of-sight (LoS) to non-line-of-sight (NLoS) conditions can be emulated. It turns out that the NN-based system cannot just compete with the LMMSE equalizer, but it also can be trained w.r.t. resilience against system parameter mismatch. We thereby showcase the conceptual simplicity of such a data-driven system design, as this not only enables more robustness against, e.g., signal-to-noise-ratio (SNR) or Doppler spread estimation mismatches, but also allows to use the same equalizer over a wider range of input parameters without the need of re-building (or re-estimating) the filter coefficients. Particular attention has been paid to ensure compatibility with the existing IEEE 802.11p piloting scheme for V2X communications. Finally, feeding the payload data symbols as additional equalizer input unleashes further performance gains. We show significant gains over the conventional LMMSE equalization for highly dynamic channel conditions if such a data-augmented equalization scheme is used.

We demonstrate an application of finding target persons on a surveillance video. Each visually detected participant is tagged with a smartphone ID and the target person with the query ID is highlighted. This work is motivated by the fact that establishing associations between subjects observed in camera images and messages transmitted from their wireless devices can enable fast and reliable tagging. This is particularly helpful when target pedestrians need to be found on public surveillance footage, without the reliance on facial recognition. The underlying system uses a multi-modal approach that leverages WiFi Fine Timing Measurements (FTM) and inertial sensor (IMU) data to associate each visually detected individual with a corresponding smartphone identifier. These smartphone measurements are combined strategically with RGB-D information from the camera, to learn affinity matrices using a multi-modal deep learning network.

5G offers high flexibility at the radio, transport and core networks to support various services of critical verticals such as connected and automated driving. At the Radio Access Network (RAN), 5G defines a New Radio (NR). 5G NR utilizes different subcarrier spacing, slot durations, modulations and channel coding schemes. This flexibility offers the possibility to support automotive services with different and demanding requirements, such as Advanced Driver-Assistance System (ADAS), cooperative driving, and remote driving. Previous studies showed that 5G NR can be configured to achieve latencies below 2 ms. However, existing studies are generally restricted to scenarios with a limited number of users and unlimited bandwidth. Therefore, it is important to analyze whether 5G NR can effectively support these services as the network scales under limited spectrum allocations. This study advances the current state of the art to demonstrate that the capability of 5G NR RAN to support advanced V2X services depends on the RAN configuration (subcarrier spacing, slot duration and error protection) and network load.

This paper proposes a stochastic model for the Global Positioning System (GPS) position errors on smartphones in urban canyons. The need to simulate such errors arises in pedestrian to vehicle communications, for example, which enable a new way to protect vulnerable road users. Studying this technology requires accurate modeling of the GPS precision on portable devices, e.g., smartphones since they share pedestrians' location information with vehicles for collision avoidance. The model is derived from and calibrated with pedestrian GPS traces collected from New York City. We show that the model produces GPS error samples with similar spatial and temporal correlation as in the collected field data.

Currently, many automated vehicle systems primarily perceive the environment from a single perspective and as a result are unable to leverage additional scene information from the viewpoint of other vehicles on the road using vehicleto-vehicle communication technologies. We study how increased data sharing can improve the perception capabilities of automated vehicles. Our methodology shares sensor measurements and objects detected by state-of-the-art deep learning networks between vehicles to increase the automated driving systems confidence of detecting objects using a 3D sensor fusion algorithm. This approach can benefit scenarios where an object may be occluded (fully or partially) or located too far away to classify accurately by a single vehicle alone.

This paper reports on our methodology and experience from a multi-year effort to cross-validate a vehicular network experiment with four hundred Dedicated Short Range Communications IEEE 802.11p transmitters through ns-3 simulations. With most of these transmitters in communication range, this represents an extremely dense wireless configuration that challenges radio and interference models. Field test and simulations were conducted in tandem and iteratively to facilitate model selection and configuration as well as to allow a detailed evaluation of simulation accuracy. We have learned that 1) results were most sensitive to parameter choices in the propagation and receiver models, with simulator default parameters not providing a good match; 2) results could, however, be significantly improved by adapting, implementing, and calibrating the propagation models and receiver models from the literature, yielding 88% accuracy (in terms of packet error rate compared to the field test) in such a complex large-scale setting; 3) the process was helpful in identifying errors both in the simulation models and in the experimental code and points to opportunities for further research.

Channel congestion is one of the major challenges for IEEE 802.11p-based vehicular ad hoc networks. To tackle the challenge, several algorithms have been proposed and some of them are being considered for standardization. Situations could arise where vehicles with different algorithms operate in the same network. Our previous work has investigated the performance of a mixed-algorithm vehicular network for the CAM-DCC and LIMERIC algorithms and identified that the CAM-DCC vehicles could potentially experience a performance degradation after introducing the LIMERIC vehicles into the network. In this work, we study whether it is possible to eliminate or bound this degradation. We propose a CBP target adjustment mechanism which controls the CBP target of LIMERIC vehicles according to vehicle density and mixing situation of the two algorithms in the network to limit the performance degradation of CAM-DCC vehicles to a desired level. The proposed mechanism is evaluated via both MATLAB and ns-2 simulations and the simulation results indicate that the performance degradation of the CAM-DCC vehicles is controlled as expected with only negligible impact on the performance of LIMERIC vehicles, which still perform similar or better than CAM-DCC vehicles.

Channel congestion is one of the major challenges for IEEE 802.11p-based vehicular networks. Unless controlled, congestion increases with vehicle density, leading to high packet loss and degraded safety application performance. We study two classes of congestion control algorithms, i.e., reactive state-based and linear adaptive. In this paper, the reactive state-based approach is represented by the decentralized congestion control framework defined in the European Telecommunications Standards Institute. The linear adaptive approach is represented by the LInear MEssage Rate Integrated Control (LIMERIC) algorithm. Both approaches control safety message transmissions as a function of channel load [i.e., channel busy percentage (CBP)]. A reactive state-based approach uses CBP directly, defining an appropriate transmission behavior for each CBP value, e.g., via a table lookup. By contrast, a linear adaptive approach identifies the transmission behavior that drives CBP toward a target channel load. Little is known about the relative performance of these approaches and any existing comparison is limited by incomplete implementations or stability anomalies. To address this, this paper makes three main contributions. First, we study and compare the two aforementioned approaches in terms of channel stability and show that the reactive state-based approach can be subject to major oscillation. Second, we identify the root causes and introduce stable reactive algorithms. Finally, we compare the performance of the stable reactive approach with the linear adaptive approach and the legacy IEEE 802.11p. It is shown that the linear adaptive approach still achieves a higher message throughput for any given vehicle density for the defined performance metrics.

We describe three in-field data collection efforts yielding a large database of RSSI values vs. time or distance from vehicles communicating with each other via DSRC. We show several data processing schemes we have devised to develop opportunistic Vehicle-to-Vehicle (V2V) propagation models from such data. The database is limited in several important ways, not least, the presence of a high noise floor that limits the distance over which good modeling is feasible. Another is the presence of interference from multiple active transmitters. Our methodology makes it possible to obtain, despite these limitations, accurate models of median path loss vs. distance, shadow fading, and fast fading caused by multipath. We aim not to develop a new V2V model, but to show the methods enabling such a model to be obtained from in-field RSSI data, without elaborate measurement design and the associated deployment cost. Finally, models based on field data allow for capturing the multiple effects of an increasing number of simultaneous V2V transceivers under typical extreme traffic scenarios.

Channel congestion is one of the major challenges for IEEE 802.11p-based vehicular ad hoc networks. Facing the challenge, several algorithms have been proposed. Two good representatives are the Decentralized Congestion Control (DCC) framework defined by ETSI and LIMERIC, a linear control algorithm. Both algorithms control message transmission rate as a function of channel load (i.e., channel busy percentage, CBP). In this work, DCC is assumed to be deployed for day one applications and LIMERIC is introduced into the network afterwards. Given such a mixed vehicular network with vehicles running either DCC or LIMERIC, we evaluate the performance of the two algorithms in the mixed scenario and study the impact of such mixed network operation on the performance through ns-2 simulations. It is observed that converting some CAM-DCC vehicles to LIMERIC vehicles will not lead to any significant performance degradation of either CAM-DCC or LIMERIC vehicles. Any observed performance degradation is small enough so that we would expect typical DSRC applications to still be feasible. Performance differences between DCC and LIMERIC exist but they can be reduced through careful selection of the specific parameters of LIMERIC and DCC, respectively.