Log In Sign Up

V2X-Based Vehicular Positioning: Opportunities, Challenges, and Future Directions

Vehicle-to-Everything (V2X) will create many new opportunities in the area of wireless communications, while its feasibility on enabling vehicular positioning has not been explored yet. Vehicular positioning is a crucial operation for autonomous driving. Its complexity and stringent safety requirement render conventional technologies like RADAR and LIDAR inadequate. This article aims at investigating whether V2X can help vehicular positioning from different perspectives. We first explain V2X's critical advantages over other approaches and suggest new scenarios of V2X-based vehicular positioning. Then we review the state-of-the-art positioning techniques discussed in the ongoing 3GPP standardization and point out their limitations. Lastly, some promising research directions for V2X-based vehicular positioning are presented, which shed light on realizing fully autonomous driving by overcoming the current barriers.


Realizing Multi-Point Vehicular Positioning via Millimeter-wave Transmission

Multi-point detection of the full-scale environment is an important issu...

Joint Frequency-and-Phase Modulation for Backscatter-Tag Assisted Vehicular Positioning

Autonomous driving (auto-driving) has been becoming a killer technology ...

OCR-RTPS: An OCR-based real-time positioning system for the valet parking

Obtaining the position of ego-vehicle is a crucial prerequisite for auto...

Integrating Geometry-Driven and Data-Driven Positioning via Combinatorial Data Augmentation

Precise positioning has become one core topic in wireless communications...

IOTA Feasibility and Perspectives for Enabling Vehicular Applications

The emergence of distributed ledger technologies in the vehicular applic...

Cooperative Multi-Point Vehicular Positioning Using Millimeter-Wave Surface Reflection (Extended version)

Multi-point vehicular positioning is one essential operation for autonom...

Sensing Hidden Vehicles Based on Asynchronous V2V Transmission: A Multi-Path-Geometry Approach

Accurate vehicular sensing is a basic and important operation in autonom...

I Introduction

Autonomous driving has been a long-term desired technology for enriching our lives by automating transportation and thereby freeing human drivers from their tasks. It will allow us to enjoy sparse time as passengers during the journey. Besides, a group of cars will cooperate to reduce traffic congestion as well as improve fuel efficiency (e.g., platooning). It is widely agreed that autonomous driving will be a disruptive technology for next-generation smart transportation. Billions of dollars in R&D investments have been made not only by major automobile manufacturers (e.g., BMW, Tesla) but also by Internet companies (e.g., Google, Baidu). With rapid advancements in multidisciplinary technologies, the state-of-the-art autonomous driving technologies are being quickly transferred from labs to our real lives. It is expected that conditional autonomous driving within limited areas (e.g., highway and rural area) will be available by , and fully autonomous driving will be realized by .

An essential operation in autonomous driving technology is positioning, namely recognizing the car’s absolute and relative positions concerning other objects such as buildings, pedestrians, and other vehicles. Autonomous driving places much more stringent requirements on positioning accuracy than other services because an error can lead to fatal accidents as exemplified by recent cases involving Tesla’s and Uber’s test cars. Specifically, vehicular positioning is a challenging task due to a wide range of requirements, including high accuracy and reliability, ultra-low latency, and cost-efficiency. Unfortunately, conventional positioning approaches based on the Global Positioning System (GPS), RADAR, and LIDAR can achieve centimeters-level accuracies only in limited scenarios. These cannot satisfy other requirements summarized below.

  • Reliability: Accurate vehicular positioning should always be available without an outage, but surrounding environments are continuously changed due to high mobility and may become unfavorable. For example, even though GPS is widely used for absolute positioning, it does not work under urban canyon environments where Line-of-Sight (LoS) links to GPS satellites are often blocked. On the other hand, the relative positioning using RADAR and LIDAR are valid only when LoS to targets are present. One solution for cases without LoS is to infer the current location from the latest positioning signal based on trajectory tracking, called dead reckoning (DR). However, DR relies on onboard sensors such as gyroscope and odometer. Its accuracy cannot be guaranteed and deteriorates as the non-LoS (NLoS) duration grows.

    Figure 1: Operation scenarios in V2X-based vehicular positioning
  • Latency: The current state-of-the-art technology, LIDAR, suffers from the long latency caused by collecting a large amount of data from scanning the surrounding environment and processing the data. Though it is capable of achieving centimeter-level accuracy and widely used for mapping, its application to real-time positioning is limited due to the scanning and processing latencies.

  • Cost: Automobile manufacturers are struggling with the high cost of autonomous driving technologies. In particular, many onboard sensors and powerful processing units are needed for accurate positioning, causing high development cost. For example, a LIDAR costs about 10,000 USD.

Recently, Vehicle-to-Everything (V2X) has emerged as a new type of vehicular communications comprising Vehicle-to-Infrastructure (V2I), Vehicle-to-Vehicle (V2V), and Vehicle-to-Pedestrian (V2P) communications [seo2016lte]. Along with the recent evolution of wireless networks termed New Radio (NR), V2X is expected to provide many new services for autonomous driving. One main direction is positioning that has not been actively explored yet. In this article, we first begin with a discussion on how V2X can help vehicular positioning. There exist a few recent works dealing with vehicular positioning in NR, especially focusing on millimeter-wave (mmWave) bands (see, e.g. [wymeersch20175g]). In this article, we aim at investigating vehicular positioning from V2X perspective covering all possible frequency bands in NR, including mmWave and micro-wave (Wave) bands. Then we introduce up-to-date standardization activities of NR’s positioning in The 3rd Generation Partnership Project (3GPP), and list a few open technical challenges. Finally, we point to several promising directions in which solutions for overcoming the current challenges can be found and summarize relevant research issues and opportunities.

Ii V2X-Based Vehicular Positioning

Though V2X designs so far have focused on wireless communications to provide reliable and efficient data delivery between vehicles, V2X has a great potential from the positioning perspective. In this section, the key advantages of V2X are first discussed in the context of fulfilling the requirements mentioned above. Then, the scenarios where V2X-based vehicular positioning finds its strengths are described.

Ii-a Advantages of V2X for Vehicular Positioning

V2X has the following distinct features for overcoming the drawbacks of the conventional technologies mentioned above.

  • Robustness of V2X channel: Recall the reliability issue of GPS, RADAR, and LIDAR where frequent outages occur in cases with LoS blockage. Consequently, GPS cannot be used in places such as indoor, tunnel, and underground, while RADAR and LIDAR may not be able to detect an object in complex surroundings. On the other hand, most frequency bands of V2X channels, which are verified to be effective in these environments, can improve the reliability of vehicular positioning.

  • Communication-aided positioning

    : An autonomous vehicle should distinguish whether targets are vehicles, pedestrians, or others for importance-aware object tracking. To this end, the car needs to estimate not only the location but also other information of a target (e.g., speed, size, and shape), which places a heavy burden on its processing units. Consequently, the resultant positioning accuracy is degraded and the latency is further increased. With V2X, on the other hand, time-invariant information like the target’s shape and size can be delivered via reliable wireless communications. Then, the processing unit can be fully utilized to estimate time-varying details, such as the vehicle’s location and speed, leading to more accurate positioning and shorter latency.

  • Infrastructure-reuse: Most signal processing techniques of positioning are similar to those of communications, e.g., detection, filtering, and beamforming. It is thus possible to reuse the existing V2X infrastructure such as base stations (BSs) and road side units (RSUs) with software only upgrading, which is a cost-effective solution.

Ii-B Scenarios of V2X-Based Vehicular Positioning

Depending on whether vehicles’ locations are communicated or estimated, two types of positioning approaches are presented. First, message-based positioning is to let the car know the position of itself or its neighbors by receiving the message containing the explicit location information via V2X communications. Second, waveform-based positioning is to estimate the locations based on the physical relations among several measurement results, termed positioning elements in this article. The positioning elements can be estimated by detecting a waveform referring to a specific signal shape already known by the vehicle. The particular positioning elements and techniques will be elaborated in Section III.

Based on the two, we introduce the operating scenarios of V2X-based vehicular positioning as illustrated in Fig. 1.

Scenario 1. The absolute or relative positions collected by GPS, RADAR, or LIDAR et al. can be transmitted to neighbor vehicles via direct V2V transmissions.

Scenario 2. The absolute or relative positions collected by GPS, RADAR, or LIDAR et al. can be transmitted to vehicles in NLoS via multi-hop V2V transmissions.

Scenario 3. A vehicle’s absolute position can be estimated by detecting waveforms transmitted by anchors, which are the infrastructure whose locations are known in advance, i.e., BSs or RSUs.

Scenario 4. The relative position of a vehicle can be estimated by detecting the waveforms transmitted by other vehicles.

Scenario 5. A macro BS broadcasts the entire positioning map within its coverage, which can be made through vehicles’ measurement reports obtained in Scenarios 1-4.

Message-based positioning is used in Scenarios 1, 2, and 5, while waveform-based positioning is used in Scenarios 3 and 4. In the past decade, the former has been considered as a standard approach for vehicular positioning, creating an active research area named cooperative positioning. With the aid of cellular and V2V communications, vehicles can obtain and deliver the sophisticated sensing data measured by various technologies including GPS, RADAR, LIDAR, or cameras. Recent advancements in cooperative positioning can be found in numerous surveys such as [alam2013cooperative]. On the other hand, it is valid only when the aforementioned technologies work well. Otherwise, vehicles should rely on the latter. Despite the importance, the latter is a relatively new direction in the area of vehicular positioning, while it has been studied in indoor and cellular localizations. Thus, we will mainly discuss waveform-based vehicular positioning.

Iii The State-of-the-Art Positioning in 3GPP

This section reviews the state-of-the-art positioning techniques that have been discussed in 3GPP NR. First, we explain key positioning elements as a preliminary. Then, we introduce the recent 3GPP standardization of positioning and describe the limitations when applied to vehicular positioning.

Iii-a Introduction to Key Positioning Elements

Depending on the means used to capture a certain physical property of radio signals, positioning elements are categorized into time-based, phase-based, angle-based, and frequency-based.

  • Time-based: Time-based elements use one basic theory of classical physics that a signal propagation speed is constant at light speed (m/sec). Time-of-Arrival (ToA) refers to the flight-time of a radio signal between an anchor and a User Equipment (UE), which can be translated into the corresponding flight-distance by multiplying . The relation relies on the assumption of anchor-UE synchronization, which is sometimes infeasible in practice. To cope with it, Time-Difference-of-Arrival (TDoA) is introduced, referring to the time difference between two radio signals’ arrivals originated from or received at different anchors. In the same vein, TDoA is equivalent to the difference of the flight-distance from the vehicle to the two anchors. Geographically, a ToA represents a circle for the UE’s possible position due to a radio signal’s isotropic propagation. On the other hand, a TDoA represents a hyperbolic curve because it represents the difference between two Euclidean distances.

  • Phase-based: A phase is another factor that constantly varies with the corresponding frequency. Phase-of-Arrival (PoA) is the phase change from the initial value, which is proportional to the corresponding flight-distance. Due to the nature of periodicity, the available range without ambiguity is the reciprocal of the concerned frequency, which is too short to use in the frequency range of NR. It is thus recommended to use Phase-Difference-of-Arrival (PDoA) defined as the phase difference between two arrival signals with different frequencies. The available range of PDoA is the reciprocal of the frequency gap, which is significantly larger than the former.

  • Angle-based: Due to the advance of signal processing techniques and the appearance of massive antenna arrays, it is possible to find out a radio signal’s propagation directions precisely, namely Angle-of-Arrival (AoA) and Angle-of-Departure (AoD). For instance, consider a signal departing from a single transmit antenna. Based on the far-field assumption, the received signals’ phase gaps at different receive antennas depend on AoA. Conversely, signals originating from different transmit antennas cause a phase gap of the received signal according to AoD.

  • Frequency-based: Due to the Doppler effect, the frequency of the received signal is shifted proportionally to the corresponding relative velocity, defined as Frequency-of-Arrival (FoA). In other words, FoA embeds the vehicle’s location and relative velocity. Similar to PoA, there also exists an ambiguity issue, making it difficult to measure a true FoA, especially when the UE is moving fast. Frequency-Difference-of-Arrival (FDoA), defined as the Doppler shift gap between two different frequencies, is preferably used to avoid the ambiguity within a practical velocity range.

Various combinations of the positioning elements can be used to infer the UE’s location (see, e.g. [Laoudias2018]). Some have been considered in NR are introduced in the following.

Iii-B Standardization of Positioning in 3GPP

3GPP NR is expected to combine multiple dimensions to improve positioning accuracy substantially by exploiting a wide range of operating frequencies (both mmWave and Wave bands) and utilizing massive antenna arrays. Recently, a new work item (WI) on NR positioning support was initiated in 3GPP Release 16 [SID-NR-Positioning], which will continue until December 2019. This WI will adrress Radio Access Technology (RAT)-dependent, RAT-independent, and hybrid methods to address regulatory policies as well as commercial use cases. Primarily, this WI’s priority will be to evaluate and specify the feasibility and scalability of the existing RAT-dependent positioning techniques, including Enhanced Cell-Identifications (ECID), Observed-TDoA (OTDoA), and Uplink-TDoA (UTDoA).

  • ECID: It is a Cell-ID based method where the UE’s position corresponds to the geographical coordinates of the serving BS obtained by using tracking-area-update or paging. To improve the accuracy, ECID uses Round-Trip Time (RTT) between the UE and the serving BS, which give a circle for the UE’s possible location. A unique positioning is possible if AoA can be measured.

  • OTDoA: It is based on the downlink TDoA measurements of Positioning Reference Signals (PRSs) from multiple BSs. The measured TDoAs are fed back to the location server via the serving BS for calculating the UE’s location. Recall that one TDoA from two BSs gives a hyperbolic curve whose two foci are the corresponding BSs’ locations. At least three BSs are required for obtaining a unique position in order to find the intersection of the hyperbolic curves. It is worth noting that the PRS receptions from more BSs lead to more accurate positioning. Multiple BSs’ PRS transmissions thus need to be separated in a time-division manner, called PRS occasion.

  • UTDoA: It is an uplink counterpart to OTDoA standardized in 3GPP release 11. To enable the TDoA measurement at BSs, a Sounding Reference Signal (SRS) is used, which is an uplink reference signal for uplink channel quality estimation. Since a UE’s transmit power is limited, its performance is worse than OTDoA. On the other hand, TDoA measurements are performed at BSs instead of UE, and the computation complexity impact on the UE is less than OTDoA. It helps UE’s energy savings and lifetime extension.

Iii-C Technical Limitations

Unfortunately, the requirements of vehicle positioning are much more demanding than those of the WI in 3GPP. For example, in the WI’s commercial use cases, horizontal and vertical positioning errors for outdoor UEs are respectively m and m, while much higher level of accurate vehicular positioning is needed such as m lateral accuracy for platooning according to [TR22.886]. We explain several technical limitations that make it challenging to fulfill the requirements.

  • Limited bandwidth: A common drawback of the above positioning techniques is the low resolution due to limited bandwidth. In principle, the bandwidth of a waveform is proportional to the time resolution; a waveform with MHz bandwidth yields nsec resolution, which is translated to m positioning error when considering the speed of light

    . Exploiting the large bandwidth of a mmWave band leads to increasing time-resolution, but its efficiency is questionable due to its severe propagation loss and limited coverage. Compressive sensing is another approach for achieving super-resolution, but its sparsity constraint is not always satisfied in practice, such as high-density vehicle environments.

  • Network synchronization: OTDoA and UTDoA rely on the critical assumption that all anchors are perfectly synchronized. However, there inevitably exist synchronization errors among anchors due to some practical reasons, e.g., signaling overhead, finite fiber links, and clock misalignment. Besides, due to network densification, more small-size BSs such as picocells and femtocells are likely to be deployed in a plug-and-play manner, making the synchronization more difficult. In the WI, the synchronization gap between BSs is considered to be up to nsec [SID-NR-Positioning], corresponding to m localization error that is unacceptable for autonomous driving.

  • Unavailable in NLoS environments: The conditions for these techniques to have a unique positioning is that the number of anchors with an LoS link should be at least the minimum (i.e., for OTDoA and UTDoA). Unfortunately, the number is random depending on the current surrounding environment and can be less than the minimum. It is vital to develop a vehicular positioning technique that can be used in a NLoS environment, called hidden vehicle positioning. In the literature, there are two kinds of approaches for hidden vehicle positioning. The first is to compensate for NLoS links’ detour distance by identifying them based on statistical techniques such as [yu2009statistical]. The second is to incorporate NLoS paths to estimate the vehicle location based on geometry relations such as [miao2007positioning]. However, their usages are limited when the ground-truth distribution is unknown or the perfect synchronization among anchors is difficult.

  • Prone to high mobility: These techniques require time for several transmissions between anchors and a UE, which is vulnerable to high mobility. In the case of OTDoA, vehicles keep moving during multiple PRS receptions, and there may exist misalignment between real and estimated positions. Besides, a vehicle’s fast velocity results in a Doppler shift, making the problem more challenging.

Figure 2: ToA and PDoA’s distance estimation errors under AWGN channels. The positioning bandwidth of ToA and PDoA are MHz and KHz, respectively.
Figure 3: PDoA’s distance estimation errors under frequency-selective channels following TDL-B channel in 3GPP NR channel model [ChannelModelingNR]. In the scheme with kHz (blue curve), we use two frequencies with the separation of resource blocks (RBs). In the scheme with

MHz (red curve), one frequency is allocated in every odd RB and

frequencies are used in total since there are RBs in MHz.

Iv Research Directions for V2X-Based Vehicular Positioning

This section aims at giving several promising directions of V2X-based vehicular positioning, which deserve consideration for overcoming the above limitations. Besides, relevant research issues are presented together to make the techniques efficient and practical.

Iv-a Use Phase-Based Elements for Band-limited Scenarios

The positioning techniques in NR rely on time-based elements (i.e., ToA and TDoA), which are discontinuous information determined by the concerned sampling interval. The resultant positioning accuracy is thus limited if the allowable bandwidth is narrow according to Nyquist sampling theory. On the other hand, phase-based elements, which are continuous information over , can embed more information than time-based ones. They can provide acceptable positioning accuracy in band-limited situations. For example, PDoA, the product of the propagation delay and the difference between the two frequencies, is valid unless the two frequency tones are the same. Noting that the frequency gap can be regarded as the bandwidth, PDoA-based positioning is likely to be more resource-efficient than time-based ones. As a result, it can be beneficial in terms of radio resource management and collision avoidance since the amount of radio resource for a single vehicle is reduced. It is verified by the simulation in Fig. 2 showing the performance comparison between ToA- and PDoA-based distance estimations under an AWGN channel. The distance error of PDoA-based scheme keeps decreasing as Signal-to-Noise Ratio (SNR) increases. On the other hand, ToA-based one has performance saturation in the high SNR regime due to the limited bandwidth. For the approach to be practical, some issues are summarized below.

  • Phase ambiguity: Recall the ambiguity issue where PDoA becomes the same in every interval of where is the speed of light, and is the frequency gap. The maximum distance estimation range is thus limited for avoiding the ambiguity. It is overcome by a hierarchical scheme based on multi-frequencies. First, the distance to a target is roughly estimated by using two adjacent frequencies, which makes it possible to reduce the estimation range. Next, it is fine-tuned by using two frequencies with larger difference within the reduced range.

  • Frequency-selective channel: As the signal bandwidth increases, more channel taps are observed, and the channel becomes frequency-selective. It hampers the accurate detection of PDoA since these multi-path signals with different propagation paths are non-coherently combined, and the phase information is distorted. Two methods are considered to address this issue. The first is spatial filtering, a kind of receive beamforming to decrease an observable range in an angular domain. It plays a role to suppress the multi-path effect but the residual multi-path components may remain. The second is to use channel estimation information. The multi-path delay profile of the concerned channel helps the cancellation of phase components contributed by selective fading [PDoA_Freq_Sel]. The effectiveness of this approach is verified by the simulation result in Fig. 3. It is shown that the usage of higher bandwidth ( MHz) can improve the accuracy of PDoA-based distance estimation despite frequency selectivity of the channel. It achieves an error of less than m, which is nearly acceptable for vehicular positioning.

Figure 4: Crossing points between two hyperbolas and between a hyperbola and an ellipse when distance estimation error exists.
Figure 5: Hybrid positioning between UTDoA and OTDoA

Iv-B Use a Hybrid Technique between OTDoA and UTDOA

Though each positioning technique has different drawbacks, it is sometimes possible to compensate them and achieve better positioning accuracy by interplaying between different techniques. One representative example is the hybrid positioning technique between OTDoA and UTDoA [HybridPos]. Recall that both OTDoA and UTDoA find an object’s position from the intersections of multiple hyperbolic curves. Unfortunately, it is unlikely to achieve accurate positioning if anchors are located on a single line, e.g., RSUs alongside a road. The corresponding hyperbolic curves tend to have similar shapes. Small distance estimation error thus results in significant positioning error (see the crossing point of the black and blue curves in Fig. 4). On the other hand, the integration of OTDoA and UTDoA can reduce such an error by introducing a wide range of different geometries. To be specific, this technique can estimate each anchor’s ToA as explained in Fig. 5, which is illustrated as a circle. Besides, an ellipse, in which the foci are two anchors, can be drawn from the accumulation of two ToAs. It is shown in Fig. 4 that finding the intersection between the hyperbolic curve and the ellipse makes the positioning technique more robust against the error from individual distance estimations. The resultant positioning accuracy can be significantly improved compared with the case of using hyperbola curves only.

It is worth mentioning that this hybrid technique can overcome network synchronization error that is the main drawback of both OTDoA and UTDoA. It is shown in Fig. 5 that the three key intermediate parameters , , do not require synchronization since these are relative time differences measured at each entity. As a result, the tight integration between OTDoA and UTDoA efficiently manages multifold issues that each standalone technique cannot address, and improve positioning accuracy without using additional radio resources. For this technique to be more productive, we summarize the following design considerations.

  • Waiting duration determination: There exists a waiting duration between OTDoA and UTDoA, denoted by in Fig. 5, which plays a pivotal role to measure individual ToAs. There are two criteria to determine this duration. Since a vehicle’s high mobility results in a significant change of the corresponding location, it is better to reduce this duration as much as possible. On the other hand, this duration should be longer than the processing and encoding times required to prepare the relevant data, i.e. and .

  • Uplink power control: As aforementioned, a UE’s uplink transmit power is controlled adequately to connect the nearest anchor. It is thus difficult to accurately estimate the ToAs of other anchors due to relatively low received SNRs. During this procedure, a UE needs increasing its transmit power for the connectivity to multiple anchors that send the counterpart OTDoA signals.

Iv-C A Multi-Path Channel Helps Hidden Vehicle Positioning

The multi-path nature of a wireless channel can open a new dimension for enabling hidden vehicle positioning [han2018sensing]. When an anchor broadcasts a waveform, its multiple replicas can be delivered to a hidden vehicle through different signal paths. It enables the car to estimate an individual signal path’s positioning elements. By the interplay of these positioning elements, we can achieve the hidden vehicle’s unique positioning. For ease of exposition, consider a 2D single-bounce scattering model where each path has only one reflection. Each signal path is thus represented as the corresponding azimuth AoD, AoA, and the scatter location. Note that all signals depart from a single origin (anchor) and arrive at a single destination (hidden vehicle). It is thus possible to form a bilateral relation between them through the multiple positioning elements. For example, consider two signal paths with AoDs and AoAs, and their TDoA as in Fig. 6. Assume that the flight distance of the first signal is given. Then, the AoA and AoD give all possible locations of the vehicle represented by a dotted line whose length is equivalent to the flight distance. The other signal path’s flight distance can be calculated by adding the TDoA. Given the two flight distances, we can make a pair of the dotted lines of which the crossing point becomes the vehicle’s location. Finally, connecting all crossing points for different pairs of flight distances makes a line on which vehicle can be located (solid black line). Adding one more signal path generates one more such a line, achieving a unique vehicle location by finding the intersection between the two.

Figure 6: An illustration of hidden vehicle positioning when two signal paths are detected with each AoA and AoD and their TDoA.

The multi-path-geometry approach has many potential advantages to overcome the limitations in Sec. III-C. First, by combining multi-path signals, it is possible to cancel out the error in each signal estimation, yielding more accurate positioning than a single-path based approach. Second, all signals are perfectly synchronized because all of them depart from one anchor in the same instant. Third, it is performed by one-way transmission from an anchor to a vehicle without a feedback channel, enabling an agile positioning service. We summarize some design considerations for practical use.

  • Insufficient number of signal paths: The feasible condition of this approach is that a certain number of signal paths should be observed (e.g., paths for the example above). However, the number of observable paths is random depending on the surroundings. A way to overcome this issue is to combine signal paths observed at different times until the sufficient number of signal paths is collected. Note that it requires the vehicle’s manoeuvre information. It is required to incorporate vehicles’ movement tracking and prediction within the hidden vehicle positioning.

  • Coexistence of multi-bounce signal paths: This approach becomes more complicated when more than two-bounce signal paths coexist. We can exclude multi-bounce signals if developing a new classification algorithm based on multiple positioning elements. Besides, it is possible to incorporate these paths as a part of the positioning framework considering higher-order geometries.

  • Waveform management: The extension to multi-vehicle positioning makes the problem more challenging because different vehicles can use the same waveform. From the positioning perspective, waveforms are considered as limited resources and should be well-coordinated to avoid such a collision. In 3GPP, only one waveform is allowed to use in a particular region, called geo-zoning or zone-based resource pool separation [3GPPR1]. A broader waveform region looks favorable at a glance, but it may bring about frequent collisions when there are many cars. Considering the tradeoff is a key to optimize the region.

Iv-D Use Backscatter Tags as Cost-Effective Anchors

Anchor densification can reduce the situations of hidden vehicle positioning while paying for the high deployment cost. Instead of typical anchors (e.g., BS or RSU), one viable solution is to deploy backscatter tags such that a reader-mounted vehicle senses nearby tags. Then, the backscatter tags feed back their IDs, corresponding to their locations, to the reader by modulating and reflecting the incident waveform. A key feature of backscatter technology is to wirelessly power many positioning tags, relieving the burden of battery charging. Besides, it helps reduce the deployment and production costs due to its small form factor and simple architecture without energy-hungry components.

We can deploy backscatter tags on a road, enabling a vehicle to know its own location by reading the tag’s ID when passing over it (e.g., [Qin2017]). However, this ID-based design has some disadvantages. First, the contact time between a tag and a reader becomes shorter as the vehicle moves faster, resulting in frequent ID reading failures. Besides, as heavy vehicles pass over the tags, they easily break down, and the maintenance cost increases. To overcome the drawbacks, we propose a new deployment plan in which backscatter tags are installed alongside the road [Backscatter_pos_2019]. If a vehicle and the tags are adequately separated, the reader’s coverage is also extended due to the cone-shape wave propagation. The resultant contact time is extended. Besides, it makes the maintenance of tags much easier by prolonging their lifetimes. There are some considerations to make it more practical.

  • Location mismatch between tags and a vehicle: The proposed tag deployment does not guarantee that the tag’s position is equivalent to the vehicle’s one. To compensate for the difference, not only the tag’s ID but also the relative position should be obtained. In other words, the joint communication-and-sensing design is essential to get both of them simultaneously.

  • New multi-antenna beamforming: Due to double-propagation of a backscatter channel, its attenuation loss is much more severe than other transmission technologies. Multi-antenna beamforming can overcome this limitation by forming a sharp beam to a specific direction. On the other hand, it decreases the reader’s coverage, and the resultant contact time becomes shorter. Therefore, a new beamforming technology should be developed to optimize the tradeoff for efficient vehicular positioning.

V Conclusion

The purpose of this article is to investigate V2X as a key technology to meet the stringent safety requirements of vehicular positioning. First, we have explained V2X’s distinct advantages over the conventional techniques and explained V2X-based vehicular positioning with operating scenarios. Second, we have reviewed the current 3GPP NR’s main positioning techniques and pointed out several technical limitations. Lastly, we have proposed attractive design directions for overcoming the limitations.



Seung-Woo Ko is an assistant professor with Korea Maritime and Ocean University. His research interests include intelligent communications and computing, and localization.

Hyukjin Chae is a professional research engineer in LG Electronics. His research interests include interference management, MIMO, V2X, and positioning. He is a rapporteur of 3GPP RAN Rel. 15 V2X phase 2 work items.

Kaifeng Han is a Ph.D. candidate in The University of Hong Kong. His research interests include future wireless networks, V2X, and autonomous-driving techniques.

Seungmin Lee is a senior researcher of LG Electronics. He has been working on 3GPP standardization and performance evaluation including relay, CoMP, HetNet, D2D, and V2X.

Kaibin Huang is an assistant professor with The University of Hong Kong. His research interests include mobile edge computing, distributed learning, and 6G systems.