Taxi demand prediction plays a significant role in assisting the pre-allocation of taxi resources to avoid mismatches between demand and service, particularly in the era of the sharing economy and autonomous driving. However, most studies have only tried to figure out the complex spatial-temporal pattern of taxi demand from historical taxi demand series, neglecting the intrinsic influences of regional functions, and failing to effectively capture the dynamic long-term periodicity. In this paper, we make two important observations: (1) taxi demand pattern varies significantly between different functional regions; and (2) taxi demand follows a dynamic daily and weekly pattern. To address these two issues, we adopt Points of Interest (POIs) to identify regional functions, and propose a novel BERT-based Deep Spatial-Temporal Network (BDSTN) to model the complex spatial-temporal relations from heterogeneous local and global features. In BDSTN, a Spatiotemporal Pattern Matching module is introduced to capture the complex spatiotemporal pattern of taxi demand while considering its dynamic temporal periodicity, and a Functional Similarity Embedding module is adopted to learn the functional similarity among all regions via POIs. To the best of our knowledge, this is the first work to use BERT-based architecture to learn taxi demand patterns, and is also the first to take functional similarity represented by POIs into consideration. Our experimental results on real-world traffic datasets in New York City demonstrate that the effectiveness of the proposed method outperforms the state-of-the-art methods, and that the efficiency of our proposed model is higher than other deep learning methods.

The safety-critical applications of vehicular ad hoc networks (VANETs) require high reliability and low transmission latency. IEEE 802.11p and IEEE 802.11bd are two standards proposed for such vehicular communication systems. In this paper, we propose an effective SINR-based model to conduct the QoS analysis of IEEE 802.11p/bd driven VANETs for safety applications. First, a semi-Markov process (SMP) model with a D/G/1/1 queue is tailored and modified for characterizing channel access behavior of IEEE 802.11 broadcast networks with deterministic message generation rate. Then, a novel approach is developed to derive the distribution of signal-to-interference-to-noise ratio (SINR) on each receiver in the VANET, given general channel fading/shadowing models as well as general node geometry distributions. Consequently, various QoS metrics in Physical/MAC/Application layer of VANETs are defined and evaluated. Compared with the existed analytic SINR-based models for broadcast VANETs, the proposed models are more general, more accurate, and are faster to solve. Finally, a new mechanism is shown to map the performance in physical layer into the QoS quantification in higher layer so that the QoS prediction of IEEE 80211bd driven VANETs is enabled. Furthermore, comparison between 802.11p and 802.11bd is conducted regarding how the stringent QoS requirements of selected safety applications in both human driving VANETs and autonomous driving VANETs are met by the two communication systems. Constructive conclusions are given on the suitability of IEEE 802.11bd for VANET critical safety services and the directions of future development.

In this paper, an analytic model is proposed for the performance evaluation of vehicular safety related services in the dedicated short range communications (DSRC) system on highways. The generation and service of safety messages in each vehicle is modeled by a generalized M/G/1 queue. The overall model is a set of interacting M/G/1 queues, one queue for each vehicle. The interaction is that the server is shared as it is the contention medium. To make the model scalable, we use semi-Markov process (SMP) model to capture the shared server's behavior from one tagged vehicle's perspective, where the medium contention and back off behavior for this vehicle and influences from other vehicles are considered. Furthermore, this SMP interacts with the tagged vehicle's own M/G/1 queue through fixed-point iteration. The proof for the existence, uniqueness and convergence of the fixed point is provided. Based on the fixed-point solution, performance indices including mean transmission delay, packet delivery ratio (PDR), and packet reception ratio (PRR) are derived. Analytic-numeric results are verified through extensive simulations under various network parameters. Compared with the existing models, the proposed SMP model facilitates the impact analysis of hidden terminal problem on the PDR and PRR computation in a more precise manner.

Whether the current IEEE 802.11p communication system meets the stringent quality of service (QoS) requirements for safety applications or not is still not very clear. Signal-to-interference-plus-noise ratio (SINR) distribution plays a primary role in quantifying the QoS as well as link capacity of IEEE 802.11p in one-dimensional (1-D) highway and two-dimensional (2-D) intersection road. Most of the analytical models based on stochastic geometry assumed ALOHA access to analyze the SINR distribution, while few of the works developed the stochastic geometry based model considering CSMA access but derived the SINR distribution utilizing statistical estimator. On the other hand, the current interference based probability analytic model for the SINR distribution limit the 1-D extension to 2-D, due to the very high numerical computational complexity at 2-D. In this paper, we propose an analytic model under more general non-homogeneous Poisson process (NHPP) node distribution, more general channel fading model (Nakagami) with path loss, and noise, for the study of QoS and capacity of VANET for BSM based safety applications in both 1-D highway and 2-D intersection road. The proposed model derives QoS and capacity of VANET BSM broadcast through evaluation of SINR distribution using probability theory and ordered statistics, which has much lower computational complexity compared with the unordered statistic model. The proposed model is validated by NS2 simulation and extended to the derivation of other QoS metrics such as packet reception probability, packet reception ratio, and broadcast link capacity, etc. The performance comparisons between 1-D and 2-D are implemented, and the QoS sensitivity analyses regarding the extension to multi-intersection as well as enlarging interference range are discussed.