Autonomous driving will rely on several safety-related connected applications that coexist with infotainment services for passenger entertainment. The simultaneous provisioning of resources to services with different quality of service requirements poses an immense challenge for future cellular networks. While most safety-related applications require low latencies, infotainment services usually necessitate a high average throughput. We propose a multi-cell, distributed predictive resource allocation framework with interference coordination based on channel distribution information to address the coexistence challenges. The approach first shifts packet transmission times in the so-called statistical look-ahead scheduling (SLAS) step, leveraging service properties. Inter-cell interference is coordinated by base station communication and a low-complexity fractional interference approximation. Lastly, packets are forwarded to an online scheduler according to the found transmission schedule. Moreover, we present a convolutional neural network to reduce the computational complexity in the SLAS step. Simulations show that the distributed approach performs very close to a central-controller solution at significantly lower computational complexity. It outperforms state-of-the-art schedulers in terms of transmission reliability and spectral efficiency.

The number of always-online vehicles continuously increases, and these vehicles will form an immense mobile sensor network. For example, cars can upload live temperature and precipitation information to enhance weather forecasting, and also transmit live cellular network measurements to the cloud. We leverage this vast amount of data, particularly the reference signal received power, to estimate the channel distribution information (CDI) for the vehicular environment. In particular, the proposed CDI maps depict the small-scale fading statistics for spatially separated regions, in contrast to the large-scale fading averages of classical radio maps (path loss and shadow fading). Our map generation framework includes a heuristic for clustering and predicts the fast-fading density per cluster via the Dirichlet process mixture model. This Bayesian nonparametric approach allows for modeling any fast-fading distribution (e.g., Rayleigh, Rice) without prior knowledge. We justify the choice of this approach by benchmarking it against other density estimation methods. Moreover, we support our assumption of local medium-term channel stationarity by measurements and show the framework’s effectiveness for anticipatory networking.

In autonomous driving, several applications like teleoperated driving, back-end status verification, or online gaming for customer infotainment rely on low-latency communication. Ideally, we can select a route that best supports the applications’ requirements before the journey. Therefore, route selection for autonomous vehicles might require in-advance latency predictions. End-to-end (E2E) latency prediction is a difficult task, especially when considering that it needs to be achieved with limited active probing due to cost constraints. We study continuous latency prediction and application feasibility assessment (in terms of meeting the applications’ E2E latency requirements), using a custom-designed deep learning model that leverages feature engineering for prediction error reduction. We provide insights into the model behavior utilizing recent advances in explainable artificial intelligence. Moreover, we present a novel model-agnostic approach based on active learning to leverage passive probing data. A pre-trained model performs certainty sampling, predicts artificial labels to enlarge the training dataset, and trains iteratively on the augmented set. The results show a 5 % reduction in mean average error for continuous latency prediction and an increase of up to 2.8 % in macro F1 score due to the use of passive probing data.

In autonomous driving, several safety-related connected applications will coexist with infotainment services for passenger entertainment. Serving the resulting set of diverse quality of service (QoS) requirements poses a tremendous challenge for future cellular networks. For example, safety-related applications require low latency, while infotainment services are associated with high throughput demands. To address the coexistence challenge, we propose a multi-cell anticipatory networking framework with interference coordination based on channel distribution information. The iterative approach first optimizes packet transmission times by so-called statistical look-ahead scheduling leveraging service properties. Interference calculus is applied for estimating the network's load in each step. Finally, packets are forwarded to an online scheduler based on the found transmission schedule. Simulations show that inter-cell interference management is crucial in provisioning the desired QoS. The iterative optimization framework offers superior transmission reliability and spectral efficiency.

Autonomous driving will rely on a multitude of connected applications with stringent quality of service (QoS) requirements in terms of low latency and high reliability. At the same time, passengers relieved of steering duty have the opportunity to enjoy infotainment services that are often associated with high data rates, e.g. video streaming. The simultaneous usage of such safety-related and infotainment services leads to diverse QoS requirements which are difficult to satisfy in current wireless networks. In an effort to address this issue, we propose a two-layer predictive resource allocation framework that leverages the services' properties and incomplete channel information. First, we optimize packet transmission times by a so-called statistical look-ahead scheduling to enhance the network's QoS and spectral efficiency based upon channel distribution information. Second, packets are forwarded to an online scheduler according to the outcome of this first optimization. Physical resources are assigned considering the services' QoS requirements and current channel state. We present a novel heuristic that performs real-time resource assignment. Simulations show that our approach has a potential for improving transmission reliability and spectral efficiency.

Cooperative Intelligent Transport Systems (C-ITS) are based on the exchange of messages between vehicles and with the road infrastructure in order to improve road safety, traffic efficiency, and travel comfort. Although the lower layers of the European C-ITS protocol stack support different wireless communication technologies, ITS-G5 is currently the preferred option for direct communication between traffic participants (i.e. without relaying the information through network infrastructure, such as a base station). In particular, the ITS-G5 specification is based on the IEEE 802.11-2012 standard, and therefore, implements decentralized radio resource management. As a result, the performance of ITS-G5 might be limited by congestion and hidden node problems. These issues can be overcome with an approach that centrally manages the radio resources, as in current LTE networks and proposed for the future 5G mobile communication standard. Both management approaches — centralized and decentralized — are compared in this paper with the focus on their ability to support cooperative awareness. Results show that the centralized approach is able to increase the number of vehicles experiencing high quality of service compared to the decentralized approach.

The efficient distribution of intelligent transport system (ITS) messages is fundamental for the deployment and acceptance of ITS applications by mobile network operators and the automotive industry. In particular, the distribution of road hazard warning (RHW) messages to distant vehicles requires special mechanisms. In this case, the combination of direct communication between vehicles and the wide area coverage provided by cellular networks might be crucial not only for reducing the data transmission costs but also for improving the timeliness of ITS information. Moreover, the application of clustering and cluster head selection mechanisms among vehicles can increase the efficiency of hybrid vehicular and cellular communication networks. This paper introduces a novel cluster head selection technique for the distribution of RHW messages, and proposes an implementation of another legacy technique that was originally intended for mobile ad-hoc networks (MANETs). This paper evaluates the performance of these techniques by the means of computer simulations in two scenarios with distinct congestion and propagation conditions. The simulation results show the potential benefit of hybrid networks compared with pure cellular transmissions, especially, if the novel cluster head selection technique is used.

We propose a novel unified radio frame structure and medium access control (MAC) protocol for low-latency and highly reliable vehicle-to-X (V2X) communications. The radio frame structure enables short latency transmission and the unified device-to-device (D2D) communication for V2X services. The unified MAC protocol simultaneously enables the cellular-assisted and ad-hoc D2D communications to enable reliable V2X services in full / partial / out-of-cellular-coverage scenario. The initial system-level simulations show promising performance results in benchmarking scenarios: Unified D2D MAC can achieve radio transmission latency below 5 ms at high reliability of 90% packet reception ratio and with high availability of coverage radius of up to 200 meter. Our on-going work is expected to provide further evaluation results of the unified D2D MAC in heterogeneous V2X scenarios.

Device to Device (D2D) communication underlaying a cellular infrastructure has attracted a great deal of attention due to its potential of enhancing capacity while reducing load and energy consumption, among other features. In turn, vehicle to vehicle (V2V) communication is envisioned as a special case of D2D, where vehicles driving closely exchange messages about the driving conditions to improve the safety on the road with little support from the base station (BS). In safety applications, however, it is essential that the communication can be maintained even if the cellular infrastructure is not available due to a coverage hole or a temporary network failure. In this paper, we present a novel technique called self-organizing time-frequency division multiple access (STFDMA) to allow vehicles to communicate efficiently in an ad-hoc manner while network assistance is not available. It has two main advantages compared to the very well known IEEE 802.11p standard: (i) It is compatible with envisioned fifth generation (5G) cellular networks, and (ii) it has a deterministic medium access delay, unlike carrier sense multiple access with collision avoidance (CSMA/CA), method used in 802.11p. This is achieved by (i) reconstructing the channel gains over frequency of one device to the neighboring ones, (ii) allocating resources in a distributed manner, and (iii) adapting the transmission to the estimated channel quality.

Device-to-Device (D2D) communications as an underlay to future 5G networks is being considered as a suitable platform for vehicular communication. Key in this regard is intelligent Radio Resource Management (RRM), which needs to provision adequate quality of service, in terms of the reliability and latency of vehicular transmissions. The classical approach building upon Channel State Information (CSI) is assumed to have high management overhead due to the acquisition of the said information. Hence, alternative RRM schemes, which rely on, e.g., location information, have also emerged. Such schemes, however, are more conservative in regards to the reuse of radio resources and have lower spectral efficiency. In this paper, we compare the performance of two RRM schemes for vehicular D2D underlay networks - a CSI-based one and a location-based one, in terms of their required measurement overhead, with implications to general schemes of the respective class.

Radio resource management plays a crucial role in the context of mobile Device-to-Device (D2D) communication with strict requirements on the quality of service. Addressing some of the faced issues we have previously defined a Location Dependent Resource Allocation Scheme (LDRAS). In this work we enhance its performance in heterogeneous radio propagation environments. To this end, we optimize the transmission parameters for D2D communication by employing sequential quadratic programming and formalize the definition of the underlying zone topology by means of hierarchical clustering. Moreover, we compare the performance of the enhanced LDRAS against a selected state-of-the-art reference scheme with respect to the reliability and availability requirements of automotive applications.

This paper provides a performance comparison of location-based and Channel State Information (CSI)-based resource allocation for device-to-device (D2D) communication. Our focus is on a system where the available cellular uplink resources are reused for the exchange of messages between vehicles in D2D underlay manner. In this context, we define a heuristic resource allocation algorithm with a spectral radius feasibility check that aims to satisfy the requirements of automotive applications. Simulations show that the developed CSI-based approach achieves higher spectral efficiency as compared to a location-based scheme. However, the gains come at a price of increased feedback overhead due to the CSI acquisition.

In this paper, we analyze the capacity of future cellular networks with a Device-to-Device (D2D) underlay reusing cellular uplink radio resources. We consider two resource reuse strategies and observe the behavior of the two-tier network with respect to the per-user transport capacity in both network layers. Our results indicate that the trend of network infrastructure densification and the prioritization of cellular communication over direct links might hinder D2D communication. Hence, careful system design is essential in order to enable additional D2D-based (e.g., automotive) applications in the next-generation cellular networks.

Cooperative intelligent traffic systems (C-ITS) will help improve the safety and efficiency of ground transportation by enabling the cooperation between traffic participants. Applications based on the C-ITS paradigm rely on vehicle-to-vehicle, vehicle-to-infrastructure, and vehicle-to-device (collectively, V2X) communication for the exchange of critical information and have very stringent Quality of Service (QoS) requirements on the reliability and availability of the communication links. In order to enable V2X communication using the envisioned device-to-device underlay of future 5G cellular networks, we have introduced a Location Dependent Resource Allocation Scheme (LDRAS) in [1]. In this paper, we enhance and extend LDRAS to multicell deployments, and quantify its performance based on extensive system level simulations. The results show that LDRAS significantly improves the QoS for future C-ITS applications, from radio resource management point of view. However, satisfying the requirements of such services necessitates further improvements of 5G networks over current 4G deployments.

Device-to-device (D2D) communication as an underlay to future cellular networks has been recently considered as an efficient cell offloading and capacity increasing solution. In this paper, we propose to use the D2D underlay as a carrier for automotive safety applications with very strict quality of service and reliability requirements. We propose a location dependent resource allocation scheme (LDRAS) for mobile D2D communications that fulfills the requirements of such services, while reducing the signaling overhead and guaranteeing a certain maximum interference level within the primary network and the D2D underlay, respectively. The former is ensured by applying persistent resource allocation to the vehicular D2D network. The latter is achieved with a spatial reuse scheme with fixed resource reservation, exploiting the localized nature of vehicle-to-vehicle communications. Initial simulation results, comparing the proposed LDRAS to a state-of-the-art radio resource management algorithm, are provided as a proof-of-concept and illustrate the benefits of our solution.

The new cellular communication standard 3GPP Long Term Evolution (LTE) promises high throughputs and low latencies, thus enabling even more bandwidth-demanding and real-time critical services for end-users. This is of particular interest for vehicle manufacturers who in the future intend to offer a huge variety of cooperative driver assistance services with different quality of service (QoS) settings. In this paper we analyze the suitability of LTE for future automotive off-board services in terms of transmission delays and reliability under various QoS settings. Our investigations are based on extensive LTE system-level simulations under different load conditions and network deployments as well as on a theoretical delay analysis. The results show that an accurate selection of the LTE QoS parameters is crucial in order to meet the delay and reliability requirements of future automotive applications, especially in high-load network conditions.