1 Introduction: Vehicular Communication Networks

In recent years, Vehicle-to-Everything (V2X) communications have been investigated as a means to support emerging automotive applications ranging from safety services to infotainment [14]. However, next-generation automotive systems, which will include advanced services based on sophisticated sensors to support enhanced automated driving applications [32], are expected to require very high data rates (in the order of terabytes per driving hour), that cannot be provided by current V2X technologies. A possible answer to this growing demand for ultrahigh transmission speeds can be found in next-generation Radio Technologies (RTs) and interfaces, such as the millimeter wave (mmWave) bands [27] between 10 and 300 GHzFootnote 1 or the Visible Light Communication (VLC) bands [3] from 400 to 790 THz. On the one hand, the extremely large bandwidths available at those frequencies can support very high data rates. On the other hand, the increased carrier frequency makes the propagation conditions more challenging, as blockage becomes an important issue since signals do not penetrate most solid materials and are subject to high signal attenuation [12]. The space and time variability of the channel quality and the need for beam alignment between moving nodes have an impact not only on the design of the physical (PHY) and Medium Access Control (MAC) layers but also on the upper-layer protocols, an aspect that has been mostly overlooked in the literature so far.

A possible way to improve the performance of vehicular communications is to make a clever use of the different type of data (e.g., road structure and positions of connected vehicles) to optimize network control. In addition, vehicles may exploit multiple RTs as a fallback in the case of short outages of the high-frequency links or limitations of existing interfaces. Furthermore, the network and transport protocols need to be adapted to fulfill the strict performance requirements of V2X communications.

In this chapter, we aim at identifying potential issues and research directions in the area of intelligent V2X networks that will make use of data for multi-objective optimizations, including extremely high-capacity and reliable information dissemination among the nodes. We start our journey into the next-generation automotive world from the description, in Sect. 2, of the automotive services and applications that require V2X communication. In Sect. 3, we focus on the enabling technologies that can support V2X data exchange and on their possible shortcomings in relation with the target application requirements. In addition to the transmission technologies, a key role in a vehicular communication system is also played by the networking protocols. Therefore, in Sect. 4, we provide a brief survey of existing MAC, network, and transport protocols for vehicular networks, discussing their possible evolutions to better support next-generation automotive scenarios. In Sect. 5, we discuss how different types of data can be obtained from the existing systems and exchanged among the vehicles and the infrastructure to better support the final applications. Finally, in Sect. 6, we address the key aspects related to the security/privacy issues in V2X communication systems and discuss the emerging protocols for privacy management and secure data dissemination. Section 7 concludes the chapter by summarizing the discussion and suggesting promising research directions in this context.

2 Requirements for Next-Generation Vehicular Communication Networks

Next-generation automotive systems are expected to provide multiple services with diverse goals and requirements. However, providing an exhaustive list of all applications that can possibly be offered through vehicular communication systems is rather difficult, considering their large number and wide variety. In the following, therefore, we focus on four “macro-applications” that, for their generality, complementarity and significance we believe are good representatives of the main types of next-generation automotive services. Although the requirements of such services are not yet fully specified, some qualifying characteristics can be outlined as follows.

  • Infotainment generically refers to a set of services that deliver a combination of information and entertainment. Infotainment requires low latency and stable throughput (especially for streaming of high-quality video contents) and the dynamic maintenance of a multicast communication (e.g., for gaming), which can be an issue. Reliability requirements are typically loose for these services.

  • Basic Safety services are typically characterized by very strict requirements. While the size of the exchanged safety messages is typically small (up to a few hundreds of bytes), latency must be very small to ensure prompt reactions to unpredictable events. V2X connections must also be very reliable and stable, due to the sensitive nature of the exchanged information and the potential consequences of a communication failure.

  • Cooperative Perception services deal with the enhancement of the sensing capabilities of a vehicle by sharing information with neighboring vehicles and infrastructures, with the final goal of extending the perception range of the driver beyond the line-of-sight or field-of-view of one single vehicle. This operation usually requires stable, reliable, and high throughput connections, due to the detailed nature of the shared contents, while some latency could be tolerated depending on the type of data contents exchanged among vehicles.

  • Platooning refers to the services that make it possible for a group of vehicles that follow the same trajectory to travel in close proximity to one another, nose-to-tail, at highway speeds. A significant amount of information needs to be shared by V2X communications. In addition to the strict latency requirement, the connection reliability and stability are also very critical.

Figure 1 provides a visual and qualitative comparison of the different requirements of the above-listed target applications. As we can see, strict reliability constraints and small latency values are common to all such applications, while stable communications should be guaranteed especially for safety-related services.

Fig. 1
figure 1

Visual representation of the features and requirements of the vehicular applications and services presented in this section

Full size image

From the above discussion, it is apparent that advanced vehicular services are expected to challenge the capabilities of current communication technologies, calling for innovative solutions. In the following, in particular, we discuss some specific requirements for the V2X communication system, called Communication Key Performance Indices (CKPIs) that go beyond the classical Quality-of-Service (QoS) metrics such as minimum required bitrate, maximum end-to-end latency and jitter, and maximum packet loss probability.

  • Range. In a vehicular scenario, the classical QoS requirements have to be associated with a spatial range. In fact, most services will set tighter communication requirements toward nearby vehicles, so that the resulting aggregate CKPI requirements are likely to be stricter in close proximity of the transmitter, and progressively more relaxed with distance.

  • Speed. The CKPIs should also account for the speed of the vehicles. In fact, stricter requirements (e.g., in terms of latency and connection stability) are usually associated to faster nodes, whose communication quality levels should therefore be monitored more closely than for slower vehicles.

  • Directionality. CKPIs may also depend on the communication direction. For example, safety services are likely to require higher transmission capacities toward the vehicles located within the range of the transmitting beam, to avoid accidents. Therefore, the space-dependent characterization of the CKPIs described above can actually be anisotropic in space, with some directions that are more demanding than others in terms of communication requirements.

  • Nodes Density. The CKPIs can also be affected by the density of nodes. On the one hand, a higher density may require a higher bitrate to maintain coordination among the cars. On the other hand, the bitrate and reliability requirements for broadcasting information may be relaxed in the presence of multiple vehicles that share the same data content.

  • Broadcast and Multicast. A significant portion of vehicular applications requires broadcast-type V2V communications (typically enabled by omnidirectional radios). While the supplemental use of directional radios (e.g., as in mmWave) could help improve the communication performance, they typically do not provide native support for broadcast. Some other services, such as platooning, may require multicast links instead. The support of both broadcast and multicast connectivity, hence, represents another CKPI for vehicular systems.

3 Enabling Radio Technologies for Next-Generation Vehicular Networks

In this section, we present features and limitations of target RTs that are expected to play a key role in next-generation automotive applications.

3.1 Dedicated Short-Range Communications (DSRC)

The IEEE 802.11p standard supports the PHY and MAC layers of the Dedicated Short-Range Communications (DSRC) transmission service. It can operate without a network infrastructure, removing the need for prior exchange of control information and thus bringing a significant advantage in terms of latency [1]. However, the throughput and delay performance can degrade as the network load increases (e.g., due to high user density), mainly because of the limited bandwidth and the “hidden node” problem. Furthermore, some V2X applications may require reliable transmissions beyond the communication range of IEEE 802.11p, which is typically limited to hundreds of meters. Moreover, the maximum data rate supported by DSRC, between 6 and 27 Mbps for each channel, may not be sufficient to sustain the transmission rates required by some next-generation automotive applications. For instance, high-resolution sensors may require more than 50 Mbps, while rates produced by cameras range from around 10 Mbps for low-resolution compressed images up to around 500 Mbps for high-resolution images [5].

3.2 Long-Term Evolution (LTE) Cellular

LTE offers ubiquitous coverage and collision-free packet transmission, but the support of vehicular communication services may still be limited. For example, access and transmission latency increase with the number of users in the cell, thus raising scalability issues. Despite the almost ubiquitous coverage of LTE, still the connection may not be always available, or good enough to satisfy the stringent reliability requirements under weak coverage (e.g., in tunnels, underground parking lots, rural areas, mountains). Finally, the maximum data rate of 4G-LTE systems is limited to around 100 Mbps for high mobility (though much lower rates are typical), which may not be sufficient to handle the potential gigabit rates that can be generated by next-generation vehicles [1].

3.3 Wi-Fi

Wireless networking based on the IEEE 802.11 standard, i.e., Wi-Fi technology, is popular and broadly available at low cost for home networks [19]. Raw data rates from 10 to 300 Mbps have proven to scale to several hundreds of concurrently active users when properly designed. However, Wi-Fi is mainly used by stationary or slowly moving indoor and outdoor users while, in a vehicular context, high mobility and link instability must be considered. Moreover, despite the high data rate, still the transmit speed may be insufficient to fully satisfy the requirements of some next-generation automotive applications. Finally, despite the huge popularity of Wi-Fi, the availability of access points that can be potentially used for V2X communications (e.g., at street corners, co-located with traffic lights, in parking lots, gas stations, cars, and so on) is still scarce, and the resulting intermittent connectivity will affect both the data rate and the latency of many vehicular services.

3.4 Millimeter Wave Bands

Communication in the millimeter wave (mmWave) bands [27] between 10 and 300 GHz is a promising candidate to support high data rates, in the order of Gbps, in line with the requirements of the next-generation cellular communication standard (5G). Moreover, the small wavelengths at mmWave frequencies make it practical to build very large antenna arrays (e.g., with 32 or more elements) to provide spatial isolation, reduce interference, increase security/privacy, and support multiplexing [12]. However, there are many concerns about the transmission characteristics at these frequencies [13]. The path loss is indeed very large and the communication range is quite limited. Moreover, mmWave signals do not pass through most solid materials, and movements of obstacles and reflectors, or even changes in the orientation of a handset, cause the channel to rapidly appear and disappear [35]. Additionally, mmWave links are typically directional to benefit from the resulting beamforming gain, requiring the fine alignment of transmitter and receiver beams and, consequently, a large overhead. Finally, dense deployments of short-range cells, as foreseen in future cellular networks operating at mmWaves, may increase the rate of handovers and reassociation events between adjacent cells, with consequent throughput degradation [21]Footnote 2.

3.5 Visible Light Bands

Visible Light Communication (VLC), whose bandwidth extends from 400 THz up to 790 THz, is an optical wireless communication technology that uses low-power Light Emitting Diodes (LEDs) not only to provide light but also to transmit data (e.g., brake signaling from car’s taillights). The large and unregulated available bandwidth (390 THz) provides attractive opportunities for many automotive applications, due to the huge achievable data rates. Furthermore, other radios can be used simultaneously with VLC, without interference. However, VLC coverage is restricted to small areas and to Line-of-Sight (LoS) links. Moreover, the limited modulation bandwidth of today’s inexpensive LEDs and the inter-symbol interference due to multipath propagation represent data transmission bottlenecks [3].

3.6 Satellite Communication

Satellite communication guarantees huge coverage areas, reaching zones that are not serviced by either landline or cellular networks. Moreover, since the cost of satellite broadcasting is basically independent of the number of receivers, the system scales very well with the number of served vehicles. Nevertheless, from a network perspective, satellite communication suffers from long delays, packet losses, intermittent connectivity, and link disruptions. Transmission also requires LoS conditions with the vehicle, limiting the accessibility to the service, in particular in dense urban areas. Finally, satellite channels are mostly broadcast and downlink, making this technology unsuitable for services that require unicast and uplink communications.

3.7 Low-Power Wide Area Networks (LPWANs)

Low-Power Wide Area Network (LPWAN) technologies may provide low power and low data rate connectivity over tens of kilometers. Furthermore, LPWAN base stations can connect a large number of devices, thus making it possible to cover wide geographical areas with a small number of base stations, significantly reducing the costs for infrastructure deployment. However, this technology offers very low data rates (in the order of tens of kilobits per second) with high latency (in the order of seconds or even minutes), thus restricting the possible employment of these technologies to noncritical vehicular services [26].

Table 1 provides a schematic and qualitative comparison of the requirements of the above-listed target RTs. As we can see, most radio interfaces ensure wide range, but with relatively high latency and small/medium throughput, while only a few technologies (i.e., mmWave and VLC systems) can provide high throughput.

Table 1 Features and requirements of the radio technologies presented in Sect. 3
Full size table

4 Next-Generation Vehicular Architecture

The demanding features of future vehicular networks, together with the limits of current and future radio access technologies and the peculiarities of upcoming wireless systems, have driven the redesign of the communication stack. In this section, we propose some guidelines for the design of next-generation MAC, network and transport layers specifically tailored to high-frequency vehicular communication systems.

4.1 Medium Access Control (MAC) Protocol Design

Medium Access Control (MAC) layer design has been extensively studied in the context of DSRC and 4G-LTE, while only a limited amount of literature has investigated solutions for other types of radios that are expected to be available in next-generation automotive systems. Conventional MAC solutions are suitable for situations in which the velocity/position of the vehicles can be accurately predicted. However, this may not be the case for V2X communication systems operating at high frequencies, mainly due to the intrinsic variability of the channel. Moreover, most recent solutions lack consideration of some important KPIs like reliability and delay. In particular, mmWave radio links require new schemes to enable vehicles and infrastructures to quickly determine the best directions to establish directional links. This functionality can be hardly supported by traditional communication protocols, which are often significantly affected by the high speed of the nodes and by the presence of frequent blockages on the propagation path.

The above discussion makes apparent the need for innovative MAC protocol design, specifically tailored to future vehicular networks, as represented in Fig. 2. This objective can be achieved by enabling multi-connectivity, thus coupling a high-frequency data plane with a lower frequency control plane, to support the required rates, while increasing the robustness of the communication [9].

Fig. 2
figure 2

Proposed solutions for MAC protocol design for vehicular networks

Full size image

The authors in [5] present a beam prediction technique based on periodical speed and position information exchanged among network nodes through DSRC messages. Using the acquired information, the system is then able to estimate the vehicle’s trajectory and derive the optimal beam orientation accordingly.

Beam design optimization is also being considered as a solution to maximize the data rate [32]. Results are consistent with the intuition that narrower beams should be used for users near the cell edge, where coverage is weaker.

In [8], a location-aided beamforming strategy is proposed to achieve ultrafast connectivity between nodes. In particular, adaptive channel estimation based on location information allows the estimation time to be substantially reduced.

Efficient beam alignment schemes can also be designed by extracting information from radar signals [24]. Simulations confirm that radars can be a useful source of side information and can help configure the mmWave V2I links.

In conclusion, although mmWave communication is a viable approach to provide high-bandwidth connectivity to future intelligent vehicles, innovative MAC-layer solutions should be engineered to overcome the limitations that prevent the direct employment of traditional communication protocols on high-frequency links.

4.2 Network and Routing Protocol Design

While the literature on network protocolsFootnote 3 for legacy vehicular scenarios is quite rich, little work exists regarding the communication performance of the network layer (especially routing) in a next-generation V2X context. More specifically, traditional routing solutions can be classified into two categories, as reported in Fig. 3: (i) topology-based routing protocols, and (ii) position-based routing protocols.

Fig. 3
figure 3

Review of network and routing solutions for vehicular networks

Full size image

Topology-Based Routing Protocols. These schemes use link information within the network to send the data packets from the source to the destination. In particular, proactive routing protocols continuously maintain up-to-date routes for all the valid destinations, thus guaranteeing low-latency packet forwarding but suffering from scalability issues. Reactive routing protocols, instead, establish the path to follow for packet delivery only when a message needs to be actually exchanged, thus saving precious bandwidth resources but increasing the latency to find a reliable route.

Position-Based Routing Protocols. These schemes do not require routing tables, but only use the position information of neighboring nodes to determine the next forwarding hop to the destination. Since those protocols are based only on local knowledge, they are considered more scalable and robust against topological changes. However, they exclusively rely on position information that may be inaccurate or unavailable (e.g., in tunnels or where the satellite signal is absent) [23], and may suffer large overheads or additional delays caused by collision and contention of the underlying MAC protocols.

In this context, the propagation characteristics and the directional nature of mmWave links bring both challenges and opportunities for routing protocol design. For instance, due to the presence of communication blockages, the shortest path connecting two network nodes (in terms of geographical or topological distance) is not necessarily the best, and may actually yield lower throughput and higher packet loss than a longer path. It is thus important to make a judicious selection of relaying nodes, for example trying to keep the number of hops to a minimum when using multi-hop communications to overcome an impaired direct path.

Recently, some works tried to design network layer protocols specifically tailored to multi-hop systems with directional antennas. In [4], the authors proposed an Optimal Geographic Routing Protocol (OGRP) that selects the appropriate multi-hop relays considering the specific features of mmWave propagation. Other solutions implement some sort of multipath routing that allows a vehicular node to establish multiple connections through different access technologies, besides using device-to-device (D2D) transmissions.

In [25], a multi-hop concurrent transmission scheme is proposed and, by properly breaking one single-hop low-rate link into multiple shorter high-rate links and allowing non-interfering nodes to transmit concurrently, the network resources can be efficiently used to improve the network throughput.

4.3 Transport Protocol Design

A relevant issue in vehicular networks is the performance analysis of transport-layer protocols, especially congestion control using the Transmission Control Protocol (TCP). In fact, traditional implementations are not suitable for high-frequency vehicular systems. First, standard slow start mechanisms can take several RTTs to achieve the full throughput offered by the mmWave physical layer, increasing the latency of the communication. Second, sudden drops in the data rate, which are likely to occur in LoS-NLoS transitions, can result in very large queuing at the nodes, dramatically increasing the packet drop probability. Third, after a retransmission timeout, even aggressive TCP protocols (e.g., Cubic) can take inordinately long to recover the full data rate [36].

As summarized in Fig. 4, one possible way to design mmWave-aware transport layer protocols is to dynamically adapt the TCP flow according to the instantaneous channel propagation conditions of the surrounding nodes, thereby reducing the congestion window size in case the path between the endpoints is obstructed [37].

Fig. 4
figure 4

Proposed solutions for transport protocol design for vehicular networks

Full size image

The hybrid and joint use of V2V and V2I communications can also ensure better QoS and transmission efficiency, especially when delivering large data contents [18]. The message may indeed be divided into several segments and delivered to multiple vehicles (i.e., to remove possible points of failure on the propagation paths), or shared among multiple infrastructure nodes, which are less affected by forwarding constraints.

Multi-connectivity can also be used to increase the overall throughput performance. In fact, while mmWaves can be exploited in the data plane to achieve the multi-Gbps rates required by next-generation automotive systems, legacy frequencies add robustness to the network thanks to the higher obstacles penetration capabilities and their inherent stability [9].

Finally, multipath-TCP, a standard that makes it possible to multiplex a TCP connection over multiple end-to-end paths, is another promising approach to improve the reliability of high-capacity networks. However, there are several issues with the traditional congestion control algorithms, in particular when coupling mmWave and LTE links [22].

To sum up, the first step toward the design of efficient transport protocols specifically tailored to next-generation vehicular systems operating at high frequencies involves the identification of the challenges that are specific to a high-mobility highly dynamic automotive environment and the potential performance pitfalls of existing V2X strategies. However, the definition of innovative transport-layer schemes is just in its infancy and therefore represents a wide-open research area.

5 Knowledge Acquisition and Distribution in Vehicular Communication Networks

New sensor technologies and advances in automotive electronics enable enhanced control systems and vehicle safety, and ease the driver’s workload [6]. Besides some “endogenous” sources of information, the driver’s experience can be further enhanced by exploiting inter-vehicle communication, which has the potential to dramatically expand the driver’s perception of the surrounding environment (e.g., eliminating blind spots and enlarging the field of vision). How to best utilize the acquired information to improve the communication capabilities of the system and better support the automotive services is still an open challenge. In this section, we discuss different aspects of this problem. First, we consider the most promising techniques for knowledge acquisition in vehicular communication networks. More specifically, we list the main type of information that can be exchanged among the network nodes. Second, we measure the usefulness of such acquired information for automotive-related applications. The results of such analysis can be used to drive the design of vehicular networking protocols. More specifically, identifying the correlations among different data, and measure the importance of each type of signal, can be of help toward a preliminary understanding of which piece of data can be more critical in providing actionable information toward network and application optimization. Third, we discuss the most promising techniques for knowledge distribution in V2X networks, to enable multi-objective optimization. In particular, we consider the synergistic exploitation of multiple RTs (either selectively, in parallel, or hierarchically) guarantees reliable, efficient, and stable exchange of data among nodes.

5.1 Knowledge Acquisition

In this section, we list some typical examples of information that can be collected by connected cars and utilized to optimize the vehicular communication networks and the supported services.

Global Positioning System (GPS) Data. Vehicles equipped with GPSFootnote 4 can collect a variety of information, including position, velocity, and acceleration, that can be exploited to improve V2X communications, e.g., by differentiating the data to be exchanged based on the vehicles position, or by supporting beam tracking for directional communications. However, GPS accuracy has a great impact on the overall communication performance, especially when directional communication is used (i.e., narrow beams are much more sensitive to position inaccuracy).

Cameras, LIDARs, and Radars. Radars currently operate in the mmWave spectrum between 76 and 81 GHz and are used for applications like adaptive cruise control, cross traffic alerts, and lane change. Although they enable accurate detection and localization of the surrounding objects, they are relatively less suitable for object recognition and classification purposes. Cameras use visible light or infrared and have been used as an enabler of road signs recognition, enhanced blind spot detection and lane departure alert, but require large amounts of data to be processed (i.e., from 100 to 700 Mbps, according to the target precision of images). Light Detection and Ranging (LIDAR) sensors make use of laser beams to generate high-resolution 3D depth images for accurate detection, localization, and recognition of the surrounding objects. However, off-the-shelf LIDARs are quite costly, and their required data rate is comparable to that of automotive cameras [5].

Traffic Conditions. Traffic conditions (and historical road traffic information) can be determined from GPS measurements and can be used to further improve the accuracy of the vehicle’s path selection, while guaranteeing more efficient route planning.

Driving Commands. Driving signals (e.g., braking, accelerating, steering, turning signals) can help the neighboring cars to adjust their paths and speed (i.e., according to the current driving conditions) to reduce the probability of car accidents and traffic congestion. The sharing of such signals among neighboring vehicles can then help improve safety services.

Environmental Conditions. Safety and efficiency of the driving experience can be enhanced by alerting drivers about weather conditions, including heavy rain, snow, sleet, fog, smoke, dust, ice, and black ice [31]. Some recent cars are able to download weather information provided by national weather-alerting systems from surrounding infrastructures and adapt their driving parameters accordingly.

Location Attributes and Regional Information. Efficient traffic regulation in proximity of certain locations, such as schools or hospitals, is one of the promising use cases of V2X communications. Reducing the vehicle speed or even diverting traffic to alternative roads, possibly limited to the most critical hours, may significantly reduce the probability of accidents and the traffic congestion close to these sensitive areas. Similar types of signaling can also be used to indicate temporary events (e.g., concerts, city marathons, political demonstrations), which are usually crowded and therefore can create complex and dense traffic situations to be handled.

Vehicle Types. Trucks may contribute to congestion more than cars, as they occupy more road space, take more time to accelerate, decelerate and negotiate turns, and obscure vision [30].Footnote 5 Information on the type of vehicles can be exchanged among neighboring cars over V2X communications to (i) alert the surrounding cars about the approaching of special categories of vehickles like large trucks, tractors, or buses, and (ii) differentiate the exchanged data based on the type of the destination vehicles (i.e., certain prohibitions apply to only one category of vehicles).

Historical Data. Any available data referred to previously obtained knowledge can be turned into experience. As an automotive node is able to recognize a specific profile (e.g., a driver, a place, a road), it can access its saved historical data and exploit these statistics, e.g., to adapt its driving decisions accordingly. Such historical data may not be available from the vehicles currently on the road, but can be stored in infrastructure servers and downloaded when required.

In Table 2, we schematically provide a list of the above-described information sources that can be utilized to improve specific automotive services and/or communication protocols, with a focus on the limitations and issues of such collected data.

Table 2 Types of data for knowledge acquisition
Full size table

5.2 How to Measure the Utility of Data

Next-generation intelligent vehicles are required to intensively download/upload/exchange/distribute information to enable fundamental automotive applications and services. Therefore, investigating the actual “importance” of shared data to assess whether and which specific sensor information is worth transmitting (i.e., with the final goal of minimizing the network utilization and still deliver valuable information to the receivers) is an open research challenge. A fundamental role in this regard can be played by machine learning,Footnote 6 which offers tools to perform a variety of operations, including the following:

  • Learn Which Features Have Major Impact on Target Applications. Artificial Neural Networks (ANNs) can be trained in an unsupervised manner to extract features from input vectors of different types of signals and provide a more compact representation of the input data, which makes it possible to reduce the amount of data to be exchanged, thus saving transmission capacity and reducing the load. Generally, the reliability of the learning process increases with the number of relevant ANN entries in the input set [29].

  • Detect Correlation Among Signals. By considering an input including several sources, a Generative Deep Neural Network (GDNN) [2] may reveal the presence of interdependencies among the readings of multiple sensors generated by vehicles in the same geographical area. The generative model can then be used to estimate the output samples from the input set. The accuracy of such predictions provides a way to measure the mutual information contained in different combinations of data.

  • Extract Information Features from General and Heterogeneous Signals. Once a GDNN has been trained with measurements related to the quality of the radio link (e.g., the strength of the received signal power, the bitrate, the error probability, the outage probability) and the model of the input data has been learned, the generative property of the GDNN can be exploited to predict the evolution of the input vector, or part of it, in future time instants (e.g., to predict the channel quality in the next slot and proactively adapt all protocol layers accordingly). Endowing GDNNs with reinforcement learning features [33] can also develop generative models that link actions (e.g., settings of link parameters) and effects (e.g., the corresponding performance metrics), thus making it possible to automatically find optimization actions tailored to the specific operational scenario, according to a self-configuration-self-optimization paradigm.

The importance of data content can also be assessed based on the cost (in terms of network resource consumption) to collect the data and to exchange it among the nodes. Which measurements and data are easier to predict and/or more useful to combine or share is, however, an open and challenging question.

5.3 Knowledge Distribution

While assessing the importance of different types of data plays a significant role in the efficient minimization of the network resource consumption, network utilization can be further optimized by a synergistic exploitation of multiple radio interfaces (with totally different propagation characteristics and features). More specifically, multi-connectivity (MC) [9] enables each vehicular and/or infrastructure node to integrate wireless technologies, including 3G, 4G-LTE, Wi-Fi, DSRC, mmWave, VLC, to support a variety of V2X services and benefit from the strengths of each radio technology, with the final goal of efficiently and reliably exchanging different types of data contents. Some relevant hybrid networking solutions include the following:

  • Selective Transmissions, in which data contents are transmitted through a single, dynamically selected radio interface. For instance, connected cars can maintain several signal paths to different infrastructures, operating at different frequencies, so that drops in one link can be overcome by switching data paths.

  • Parallel Transmissions, in which data contents are duplicated and sent over different types of radios to add redundancy, making the message delivery more robust, but using more communication resources.

  • Hierarchical Transmissions, in which a specific technology is used to provide a basic level of service, while different types of radios/paths are exploited to deliver supplemental information to improve the QoS of designated applications.

However, how to implement efficient multi-connectivity systems on next-generation connected cars is still an open issue. In particular, among the challenges that need to be addressed, the definition of an intelligent network selection mechanism, driven by a distributed or centralized/cloud-assisted decision process, must be engineered, to allow high-quality V2X applications to meet their requirements.

6 Emerging Protocols for Privacy Management and Secure Data Access Control in Vehicular Networks

Like any other computing system, vehicular communication networks can be plagued by vulnerabilities: connected nodes must thus be designed with security in mind, in order to limit the adversaries’ ability to endanger vehicle operations, as well as driver and passenger safety. Investigating the main security/privacy issues related to vehicular communication systems and designing protocols and techniques for privacy management and secure data dissemination are therefore important research topics.

6.1 Security Concerns in Vehicular Networks

One of the most serious threats for security in next-generation vehicular networks originates from the tens of electronic control units (ECUs) that cars will incorporate. A solution may come from consolidation, integration, and virtualization of ECUs, with the final goal of reducing the total number of electronic components and increasing the number of functions and the complexity of the software.

However, the attack surface of future automotive systems extends beyond the car itself, touching most in-vehicle systems and an increasingly wide range of external networks. The authenticity and integrity of data transmitted across networks can be improved by providing secure storage systems for key exchange and encryption, to protect against unauthorized software or firmware updates [16]. Moreover, enhanced security mechanisms in which cars will connect to smart infrastructures (e.g., toll roads, gas stations) without disclosing personally identifiable information should be developed.

Lack of sufficient bus protection is another relevant security-related concern. In fact, the Controller Area Network (CAN) bus lacks the necessary protection to ensure robust data integrity [15]. Messages on the CAN-bus are not protected by any Message Authentication Code or digital signature and can be read by other nodes that can physically access the bus.

Finally, protection of data as it moves through the cloud and to data centers is another fundamental security feature that must be provided by transportation suppliers. Reliable automotive driving experience and connected communication capabilities can be supported by optimized data encryption and by guaranteeing embedded security features in the hardware of cars [16].

6.2 Emerging Protection Mechanisms for Vehicular Networks

Recently, countermeasures have been developed to face the increasingly threatened security in next-generation connected cars. As summarized in Fig. 5, a list of emerging protection mechanisms for vehicular networks includes the following.

Fig. 5
figure 5

Proposed protection mechanisms for vehicular networks

Full size image
  1. 1.

    Network partitioning: Security can be achieved by slicing the network, one partition being responsible for the safety-critical ECUs, while the other providing “comfort” functions [17].

  2. 2.

    Secure identification and authentication: Effective software protection can be guaranteed by securely implementing authorization functions in a trusted environment [34]. Cryptography solutions can also be implemented to allow each counterpart to verify the claimed credentials.

  3. 3.

    Super ECUs: With the increasing complexity of vehicular networks, one trend is to integrate several different applications on one (more powerful) ECU. However, low-cost devices are not able to run most of these complex operating systems.

  4. 4.

    Universal rules: Malicious attacks can also be theoretically prevented with good programming practices and by following the existing security recommendations protocols. However, ECUs usually come from different manufacturers, having potentially different protection specifications. The definition of universal defense mechanisms is therefore essential to enable secure transmissions.

  5. 5.

    Secure boot: This mechanism checks the digital signature of the software, prior to execution [7]. If an asymmetric algorithm is used, the public key has to be secured only against manipulation, but not against extraction. For both types, hardware support for the key storage is necessary.

  6. 6.

    Attestation-based security architecture: By comparing the result of specific hash functions with a list of authorized hashes, only successfully validated ECUs will be able to exchange symmetric keys for further encrypted communication [20].

  7. 7.

    Redundancy of sensors: The source of the sensor data is often not properly protected, and hence the signals might still be forged. One standard approach is to use redundant sensors and authentication checks in the ECUs. In the ideal case, there are two or more sensors measuring the same physical quantity (e.g., speed) in different ways, and a cross-check ensures the plausibility of the data.

Despite the increasing efforts of the automotive industry, there are still many security-related challenges to be considered in the near future. In general, an overall standardized approach to security, accepted by industry and legislation, is still missing and is therefore a challenging research topic.

7 Conclusions

In this chapter, we highlighted the challenges raised by next-generation automotive services, with reference to the design of the communication protocol stack.

In general, in order to compensate for the increased isotropic path loss experienced at higher frequencies (i.e., at mmWaves), next-generation automotive communication systems must provide mechanisms by which the vehicles and the infrastructure determine suitable directions of transmission to exchange sensory information. In this context, the design of enhanced communication protocols (i.e., at the MAC, network, and transport layers) is fundamental to meet the requirements of next-generation V2X services. In particular, the performance of intelligent vehicles in highly mobile mmWave scenarios strictly depends on the specific environment in which the vehicles are deployed, and must account for several automotive-specific features such as the vehicle’s speed, the beam tracking periodicity, the node density, and the embedded antenna configuration.

Moreover, network resource minimization is another important issue for future intelligent vehicular systems. One possible way to achieve efficient communication is through synergistic orchestration among the multiple interfaces that are expected to be integrated in future intelligent vehicles, and by measuring the importance of data contents.

Security is another key concern for automotive networks. In this chapter, we analyzed the most serious security concerns and threats in next-generation connected cars and surveyed the most recent emerging protection mechanisms for secure data access control in vehicular networks.

Most of these research challenges, as well as many others, are still largely unexplored, so that additional investigation is needed toward the design of fully autonomous driving cars.