Dynamic Selective Positioning for High-Precision Accuracy in 5G NR V2X Networks
DDynamic Selective Positioning for High-PrecisionAccuracy in 5G NR V2X Networks
Abdurrahman Fouda, Ryan Keating and Amitava Ghosh
Nokia Bell [email protected], { ryan.keating, amitava.ghosh } @nokia-bell-labs.com Abstract —The capability to achieve high-precision positioningaccuracy has been considered as one of the most critical re-quirements for vehicle-to-everything (V2X) services in the fifth-generation (5G) cellular networks. The non-line-of-sight (NLOS)connectivity, coverage, reliability requirements, the minimumnumber of available anchors, and bandwidth limitations areamong the main challenges to achieve high accuracy in V2Xservices. This work provides an overview of the potential solutionsto provide the new radio (NR) V2X users (UEs) with highpositioning accuracy in the future 3GPP releases. In particular,we propose a novel selective positioning solution to dynamicallyswitch between different positioning technologies to improvethe overall positioning accuracy in NR V2X services, takinginto account the locations of V2X UEs and the accuracy ofthe collected measurements. Furthermore, we use high-fidelitysystem-level simulations to evaluate the performance gains offusing the positioning measurements from different technologiesin NR V2X services. Our numerical results show that theproposed hybridized schemes achieve a positioning error ≤ ≈
76% availability compared to ≈
55% availability whentraditional positioning methods are used. The numerical resultsalso reveal a potential gain of ≈
56% after leveraging the road-side units (RSUs) to improve the tail of the UE’s positioning errordistribution, i.e., worst-case scenarios, in NR V2X services.
Index Terms —3GPP Rel-16, 5G new radio, DL-TDOA, GNSS,hybrid positioning, V2X.
I. I
NTRODUCTION
Localization operations have been introduced in new radio(NR) vehicle-to-everything (V2X) services to estimate eitherthe absolute coordinates of a vehicle or its relative position tothe surrounding objects. It is of vital importance to provide NRV2X users (UEs) with stable and high-precision positioningaccuracy because a small error may lead to a fatal accident. 5GAutomotive Association (5GAA) has defined the positioningaccuracy as a service level requirement (SLR) that ranges froma few centimeters to a couple of meters based on the natureof operation of each V2X use case [1]. The 3rd GenerationPartnership Project (3GPP) has introduced native positioningsupport in NR for handheld and indoor industrial use cases [2].For V2X use cases, sidelink in NR has been designed duringRel-16 without specific positioning support [3].Generally, the proposed NR Rel-16 positioning techniquesfor handheld and indoor use cases are divided into (1) tem-poral methods, e.g., downlink time difference of arrival (DL-TDOA), (2) angular methods, e.g., downlink angle of departure(DL-AoD), and (3) hybrid schemes [4]. It is expected thatpositioning enhancements to support positioning for V2X UEs, i.e., on-board units (OBU) in vehicles, may continue in Rel-17 and beyond. The potential positioning solutions for NRV2X services can be categorized into RAT-dependent, i.e., NRbased, and RAT-independent techniques. In the former, V2XUEs use their cellular connections and the known locations ofbase stations (gNBs) for positioning using several techniques,e.g., DL-TDOA. In the later, V2X UEs exploit their satelliteconnections along with the known locations of the satellites forpositioning using global navigation satellite system (GNSS)positioning techniques, e.g., code-based GNSS, PPP-RTK [5].GNSS-based positioning has been considered as a promisingsolution in NR V2X services given the support of high-precision positioning, and the availability of many satellites.However, the limitation of the line-of-sight (LOS) connectionin many scenarios, e.g., dense urban environments, tunnels andunderground parking, is among the main challenges to achievehigh accuracy [1]. On the other hand, perfect network syn-chronization, support of high mobility, reliability requirements,the minimum available number of anchors, and bandwidthlimitation are among the main challenges to provide high-precision positioning using RAT-dependent techniques [6].The fusion of positioning measurements from differenttechnologies has emerged as a potential solution to improvethe positioning accuracy in V2X services (see, e.g., Section7.2.3 in [7] and references therein). However, the mutualdependence between the deployment environments and theaccuracy of the collected measurements is one of the mainchallenges to achieve high accuracy using these hybridizedtechniques [8]. Essentially, fusing two groups of accurate andpoor measurements will not lead to an improvement in theoverall accuracy. Hence, a new selection method is neededto determine when to have a near-optimal bias toward onetechnology over the other based on the locations of V2X UEsand the accuracy of the collected measurements.A least squares-based method for hybridization of GNSSand RAT-dependent measurements was presented in [9], [10].However, the authors did not discuss how the optimal weightscan be determined to have a specific bias for one positioningtechnology over the other. To the best of our knowledge, noneof the prior art discussed the selection between different po-sitioning technologies to improve the positioning accuracy inNR V2X services. In this paper, we propose a novel selectionmethod; namely, selective positioning based on neighboringToA variance (SPNTV) to dynamically switch between theRAT-dependent and GNSS positioning methods. a r X i v : . [ ee ss . SP ] F e b ig. 1: Different positioning modes for NR V2X services .Our numerical analysis shows that the proposed SPNTVachieves a positioning error ≤ ≈
76% availabilitycompared to ≈
51% and 58% availability when GNSS and DL-TDOA measurements are solely used, respectively. We alsoevaluate the performance of a weighted MAP-based fusionmethod and show the advantage of using the proposed SPNTVover the fusion method in several V2X use cases. Further, wepresent enhanced versions of the above hybridized positioningmethods that allow V2X UEs to dynamically switch betweenpositioning schemes (not only different technologies) based ontheir locations. Finally, we demonstrate the performance gainsof utilizing the road-side units (RSUs) to improve the tail ofUE’s positioning error distribution, in NR V2X services.II. N
ETWORK A RCHITECTURE AND S IMULATED E NVIRONMENT
In this paper, we use a two-tier cellular network to analyzethe positioning performance of two NR V2X use cases. Inthe first use case, we assume that V2X UEs are distributedrandomly at uniform in a six-lane urban road. In the secondone, we assume that a tunnel section is added to the urban road.In that, V2X UEs are divided into two equal groups, one insidethe tunnel and another outside the tunnel. As shown in Fig. 1, aV2X UE can use several modes for positioning estimation. Inmode 1, V2X UE fuses the GNSS and cellular measurementsto perform positioning. Mode 2 represents a single technologypositioning method, in which, V2X UE uses only GNSSmeasurements to perform positioning. Similarly, V2X UE usesonly cellular measurements to perform positioning in mode 3.
A. NR TDOA-based Positioning
We focus on DL-TDOA (as a cellular positioning method),but the proposed algorithms are not limited to only thismethod. A new reference signal for positioning (DL PRS)was introduced in Rel-16, which we use in this work forpositioning estimation [4]. In particular, V2X UEs measurethe reference signal time difference (RSTD) between DL PRSfrom different transmission points, i.e., gNBs and RSUs, toperform positioning [6]. The V2X UEs use the legacy LTEpositioning protocol (LPP) to exchange location informationwith the location server. A trilateral estimation algorithm isthen used by the location server to estimate the positions ofV2X UEs given the known locations of gNBs, RSUs, and theRSTD measurements. It is worth mentioning that we assumethat DL-TDOA is supported in both NR V2X UE-based (i.e.,UEs perform positioning estimation) and network-based mode. In general, the algorithms and solutions we describe in thispaper apply to both UE-based and network-based positioning.
B. GNSS-based Positioning
GNSS positioning relies on receiving signals from multiplesatellites at V2X UEs and making pseudo-range measurementsfrom these received signals. Based on the known locations ofthe satellites and the pseudo-range measurements, the UEs canestimate their locations. Let S = { , . . . , S } , U = { , . . . , U } denote the sets of GNSS satellites and V2X UEs, respectivelywhere, e.g., the cardinality of S is | S | = S . The GNSS pseudo-range measurements of u th V2X UE is given by [9]: ˜ d u = d u + e u ∀ u ∈ U , (1)where d u = (cid:0) d u,s : s ∈ S (cid:1) with d u,s represents the distancebetween u th V2X UE and s th satellite and is given by d u,s = (cid:107) p u − p s (cid:107) , where p u = [ x u , y u , z u ] (cid:62) and p s = [ x s , y s , z s ] (cid:62) are the positions of u th V2X UE and s th satellite, respectively.Similarly, pseudo-range error between u th V2X UE and GNSSsatellites is given by e u = (cid:0) e u,s : s ∈ S (cid:1) , where we model e u,s as a Gaussian distribution with N (cid:0) , (cid:62) M σ (cid:1) , σ = (cid:2) σ , . . . , σ M (cid:3) (cid:62) represents the vector of user equivalent rangeerror (UERE) variances and M denotes the M -dimensionalall-ones vector. We define UERE variance error based on theLOS status between u th V2X UE and s th satellite ρ u,s as: σ = [ σ o , σ i , σ t , σ n ] ρ u,s = 0 , [ σ o , σ i , σ t , σ n , σ m ] ρ u,s = 1 , (2)where σ o , σ i , σ t , σ n , and σ m denotes the variance of the GNSSclock, residual ionosphere, troposphere, additive white Gaus-sian receiver noise, and urban multipath error, respectively.III. P ROPOSED H YBRIDIZED P OSITIONING T ECHNIQUES
In this paper, we jointly utilize the positioning measure-ments collected from different technologies to improve theachievable accuracy in NR V2X services. In particular, wepropose two hybrid positioning schemes; namely, SPNTVand weighted-MAP based fusion. In the former, V2X UEsutilize the ToA measurements to dynamically switch betweendifferent technologies based on the estimated accuracy ofthe collected measurements. In the latter, V2X UEs fuse thepositioning measurements collected from several technologiesto perform positioning. We assume a V2X UE can performboth GNSS and DL-TDOA positioning in a UE-based mode.
A. Proposed SPNTV Algorithm
In this section, we discuss how the estimated ToA mea-surements can be used to define the selection criteria of theproposed novel SPNTV scheme. Essentially, V2X UEs usethe received DL PRS from different cells to calculate theestimated ToA. We use Monte Carlo system-level simulationsto study the performance characteristics of the estimatedToA measurements from the serving cell and the closest β neighbors. In doing so, we consider modified urban macro(UMa) and micro (UMi) scenarios, in which, V2X UEs areig. 2: TOA measurements of V2X UEs.randomly distributed over l lanes in a two-tier cellular networkfollowing the same mobility modeling in [11].We consider two use cases for the road environment in V2Xservices; urban road, and urban road with a tunnel, in which,V2X UEs suffer from penetration loss and NLOS blockagewhile being inside the tunnel. The channel realizations andV2X UE locations are randomly updated using the MonteCarlo simulations (see Section IV). To this end, V2X UEsare classified into two classes based on the estimated posi-tioning error using the DL-TDOA technique. In that, class A represents V2X UEs with an estimated positioning accuracy ≤ (cid:15) , where (cid:15) denotes the absolute positioning accuracy targetof at least 95% of V2X UEs according to the NR V2X SLRsin [1], [12]. On the other hand, the V2X UEs with an estimatedpositioning accuracy > (cid:15) are represented by class B .Fig. 2-a depicts the estimated ToA measurements with theserving cell and the first β arriving neighbors to the class A ofV2X UEs, where β = 5 . It shows that the ToA measurementsdecrease as the V2X UE becomes closer to its neighbor cell.Similarly, Fig. 2-b shows the same trend with the class B ofUEs. Figs. 2-a and 2-b reveal that the ToA measurements of thefirst arriving pair of neighbors have a high variance when V2XUEs suffer from a positioning error > (cid:15) . It is worth noting thata consistent high variance between the ToA measurements ofthe closest pair of neighbors implies a long-term poor DL-TDOA positioning performance.We utilize the above findings and introduce a new parameter ζ = t − t , where t i is the ToA of the i th arriving neighborPRS in time, and ζ denotes the time difference between theToA measurements of the closest pair of neighbors to a V2XUE. In that, V2X UE uses the first arriving pair of ToA asthe measurements of the closest pair of neighbors. V2X UEsuse the dependency between ζ (i.e., the variance of neighborToA measurements) and the estimated accuracy of the DL-TDOA to decide whether to use the cellular connection forpositioning or to switch to another technology. We implementsuch dependency in a binary fashion, in which, V2X UE onlydecides to use DL-TDOA technique for positioning if ζ < η ,where η denotes a triggering event to switch between differentpositioning technologies. We initially introduce η to be the th percentile of ζ A = ( ζ a : a ∈ A ) , where ζ a represents thedifference between the ToA measurements of the closest pair of neighbors to a class A V2X UE. In this paper, we use theGNSS as an alternative technology that V2X UEs can switchto, based on the estimated measurements of ζ A . The proposedSPNTV scheme is summarized in Algorithm 1. Algorithm 1
Dynamic selection between GNSS and DL-TDOA using the proposed SPNTV scheme. Inputs base station and GNSS locations. Initialization η = 50 th % -tile of ζ A , where ζ A is computedvia simulation campaigns. Start V2X UE measures DL PRS from both serving and neigh-boring cells and calculates ToA. Calculate ζ = t − t . if ζ ≥ η then V2X UE switches the GNSS receiver on and obtains anew GNSS positioning fix. Final V2X UE location is determined using only GNSS. else V2X UE switches the GNSS receiver off.
Final location is determined using only DL-TDOA. end if
Refine selection parameter η and goto Start .It is worth mentioning that the predefined selection thresh-old η may be dependent on the scenario, e.g., small cell,and inter-site distance. Also, η can be either computed viasimulation campaigns (see Section IV for an example) ordetermined by test V2X UEs in the field. Alternatively, areference V2X UE can learn η using machine learning overtime. For example, by comparing relative DL-TDOA or GNSSpositioning accuracy with a reference location and determining η which leads to the highest positioning accuracy. The V2XUE may also share the determined η with nearby V2XUEs via sidelink to further refine the selection. The entireprocedure may be repeated periodically to update the selectionof technology, and the periodicity may be based on either thevelocity of V2X UE, e.g., update faster for highly mobile UE,or the absolute value of ζ , e.g., for values close to η updatefaster. It is also worth noting that while the method is describedfor DL-TDOA it could also apply to multi-cell round trip timeor other future timing-based positioning methods in NR. B. Weighted MAP-based fusion Algorithm
In this section, we present the implementation of a weightedversion of the maximum a posteriori (MAP) estimator to fuseboth DL-TDOA and GNSS measurements. Compared withSPNTV, the MAP fusion algorithm uses both technologiesfor positioning, instead of selecting just a single technology.The MAP estimator initially presented in [13] for DL-TDOAmeasurements is used as a baseline. Given the known locationsof gNBs and satellites, V2X UEs utilize the MAP estimatorfor positioning. It is worth noting that the same estimator isused for positioning estimation using either the DL-TDOAor GNSS measurements only (see Sections II-A, II-B). Let G = { , . . . , G } and C denote the sets of gNBs and overallABLE I: Simulation parameters. Parameter Value
Layout 3GPP UMi, 57 cells, ISD = 200m [10].Channel model 3GPP cellular, NTN SCM for 0.5-100 GHz[14], [15].GNSS constellation 11 NGSO satellites at 1500km [15].Carrier bandwidth Cellular: 100 MHz at 4 GHz.GNSS: 24.5 MHz at 1.5GHz.NR numerology 30 KHz subcarrier spacing.Reference signal PRS, comb-6, comb-offset=0.Mobility settings Option A, speed=140 km/h, 6 lanes,density = 600 V2X UEs/km [11].Antenna geometry × , SU-MIMO, MMSE receiver.Penetration loss 20 dB (in tunnel use case).SPNTV settings β = 5 , (cid:15) = 10 m, ζ A = 16 , , , , ns.MAP setting w u,s = { e − , e − } , µ a = 0 , w u,g = { e − , e − } , σ a = 50 . anchors, respectively, where C = S ∪ G and C = S + G .The vector of fused measurements at u th V2X UE is givenby m u = (cid:0) m u,c : c ∈ C (cid:1) , in which, m u,c = ˜ d u,c if c ∈ S and m u,c = ˜ t u,c if c ∈ G where ˜ t u,c is the estimated ToAmeasurement between u th V2X UE and c th gNB. The MAPestimator is given by: (˜ p u , ˜ τ ) = arg max p u ,τ (cid:89) c ∈ C w u,c + 1 (cid:0) (cid:107) p u − p c (cid:107) + (cid:15) (cid:1) (cid:112) πσ c exp (cid:34) − σ c (cid:18) m c,u − (cid:107) p u − p c (cid:107) c − τ − µ c (cid:19)(cid:35) , ∀ u ∈ U , (3)where ˜ p u , ˜ τ , (cid:15) and c represent the estimated transmit time(unique for all anchors), the estimated position of V2X UE, anoffset to prevent numerical instability, and the speed of light,respectively. w u,a denotes the bias toward GNSS or DL-TDOAmeasurements. Finally, the measurement estimation error of c th anchor follows a Gaussian distribution with N ( µ c , σ c ) .IV. N UMERICAL A NALYSIS
In this section we investigate the achievable accuracy of theproposed hybridized schemes for a Rel-16 PRS at frequencyrange 1 (FR1) with bandwidth of 100 MHz. It should be notedthat the proposed positioning methods also apply to Rel-17and beyond where the accuracy could be greatly improveddue to PRS enhancements or wider BW, e.g., using carrieraggregation and millimeter-wave (mmWave) frequency band.For single technology-based positioning, we use the DL-TDOA and GNSS measurements given the known locationsof gNBs and satellites, respectively. Furthermore, we fuseboth measurements and use a weighted MAP-based fusionalgorithm for positioning (see Section II). We compare theperformance of the proposed SPNTV method with each one ofthe previous positioning techniques in two V2X UE droppinguse cases; namely, urban road and tunnel use cases. Fig. 3: MAP-based positioning accuracy.
A. Urban road use case
In this use case, V2X UEs are randomly distributed in asix-lane urban road and a two-tier cellular network followingthe simulation settings in Table 1. We assume that no RSUsare deployed in the evaluations of this use case. Alternatively,V2X UEs use their connections with the macro gNBs for theDL-TDOA positioning. Fig. 3 shows that the positioning errorof the DL-TDOA method outperforms that of the GNSS foralmost 60% of the V2X UEs. However, the GNSS method sig-nificantly outperforms the DL-TDOA when V2X UEs sufferfrom a poor cellular performance. In addition, Fig. 3 showsthat the positioning accuracy of fusion algorithm outperformsthat of the GNSS by almost .3 m for at least 55% of V2XUEs and maintains a decent tail behavior when class B V2XUEs suffer from poor DL-TDOA performance. Fig. 3 alsoreveals that the proposed SPNTV algorithm defines the bestpositioning technology to be used based on the V2X UElocation and maintains a long-term high accuracy for bothclasses of V2X UEs.In particular, the positioning accuracy of the proposedSPNTV scheme outperforms that of the GNSS and the fusionmethod by almost .6 m and .3 m, respectively. Fig. 3 alsoshows that the proposed SPNTV and fusion schemes achievea positioning error ≤ m with 76% and 74% availabilitycompared to 65% and 61% availability when the GNSS andDL-TDOA measurements are solely used, respectively. It isworth noting that the above findings are the same if anotherestimator (e.g. least squares) is used for positioning. However,we use the MAP estimator (in all of the above methods) giventhat the positioning performance of the MAP estimator outper-forms that of any other estimator that has been presented in theliterature. Our numerical analysis reveals that the positioningperformance of the proposed SPNTV follows that of the DL-TDOA for the class A of V2X UEs (i.e., when the accuracyof DL-TDOA is ≤ (cid:15) ), as shown in Fig. 4-a. On the otherhand, Fig. 4-b shows that the SPNTV decides to use the GNSSmeasurements when the DL-TDOA performance significantlydeteriorates. In other words, SPNTV method chooses the bestpositioning technology based on the location of V2X UEs.ig. 4: Positioning accuracy of two classes of V2X UEs.Fig. 5: Fine tuning the parameter η for the proposed SPNTV.Fig. 5 shows the advantage of having self-inherited flexi-bility in the proposed SPNTV algorithm to choose the bestpositioning technology. In that, the selection threshold η isfine-tuned to have more bias toward a specific technologyover the other. Essentially, decreasing η adds more biastoward using the GNSS measurements for positioning. Asshown in Fig. 5, the positioning accuracy of the proposedSPNTV algorithm mimics that of the GNSS when η = 16 ns. In other words, the majority of V2X UEs decide to usetheir GNSS measurements for positioning. Similarly, mostof V2X UEs use the cellular measurements when η = 220 ns. In contrast, it is a hard optimization problem to find theoptimal weights in the MAP-based fusion algorithm, giventhe mutual dependency between the accuracy of the fusedmeasurements and the deployment environment. Generally,fusing two groups of accurate and poor measurements will notlead to an improved accuracy unless there is a specific biastoward one group over the other. Hence, having a selectioncriterion (which is represented by η in the proposed SPNTV)to switch between different technologies is considered as aninstrumental factor to improve the positioning accuracy of anyhybridized positioning algorithm for practical deployments. B. Tunnel use case
In this use case, a tunnel section is added to the urbanroad such that V2X UEs are divided evenly into two groups,one inside the tunnel and another one outside the tunnel.Essentially, V2X UEs suffer from weak satellite and cellular Fig. 6: Performance of single-technology positioning method.signal coverage inside the tunnel due to the penetration lossand the lack of LOS connections. Thus, providing V2X UEswith a long-term high positioning accuracy is one of the mainchallenges in such scenarios. It is worth mentioning that theemergency brake warning is one of the typical traffic safety usecases in V2X tunnel scenarios. In that, a V2X UE is expectedto apply a brake while approaching/leaving the tunnel. Hence,it is of paramount importance that V2X UEs can achieve highpositioning accuracy anytime and anywhere in these scenarios.It is also clear that relying on GNSS solutions alone will notbe sufficient. We utilize the proposed hybridized positioningalgorithms in Section III and evaluate the positioning accuracyin this use case. Further, we investigate the potential gains ofadding RSUs inside the tunnel to provide the V2X UEs witha long-term high positioning accuracy. In that, V2X UEs usetheir connections with the gNBs and RSUs for DL-TDOAmeasurements outside and inside the tunnel, respectively.Fig. 6-a shows a severe performance degradation in GNSSpositioning at almost 80% of V2X UEs due to the tunnellosses. On the other hand, Fig. 6-b shows that the performancedegradation in cellular positioning is smaller than that of theGNSS. It also shows that adding two RSUs (at the tunneledges) improves the DL-TDOA performance to mimic thatof the baseline scenario, i.e., urban road use case without atunnel. In this regard, Fig. 7 demonstrates the robustness of theproposed SPNTV method against tunnel losses. Essentially,the proposed SPNTV ensures providing V2X UEs with highpositioning accuracy on a continuous-time basis as they movein and out of the tunnel. It also shows that the fusion algorithmsuffers from a performance degradation of almost 2 m for morethan 85% of V2X UEs compared with the urban road use case(see Fig. 3) due to the tunnel losses. In other words, the fusionalgorithm is more susceptible to any accuracy degradation ineither one of the positioning technologies (cellular or GNSS).This is because it is hard to fine-tune the fusion method tohave more bias to one technology over the other. On the otherhand, Fig. 7 reveals that, with a proper configuration of theSPNTV algorithm, it is possible to maintain high accuracy foralmost 80% of V2X UEs despite the tunnel losses (see Fig. 3for comparison with the SPNTV performance in an urban roaduse case). This can be achieved by adding more bias towardthe cellular measurements, i.e., η = 125 ns.ig. 7: Overall positioning error in the tunnel use case.Fig. 8: Positioning performance with the RSUs.It is worth noting that a significant increase of the biastoward the cellular measurements will result in degrading theSPNTV performance to mimic that of the DL-TDOA. Tosolve this problem, we evaluate the performance of enhancedversions of the hybridized positioning schemes presented inSection III. In that, V2X UEs use the DL-TDOA methodinside the tunnel and either the SPNTV or the fusion algo-rithm outside the tunnel. As shown in Fig. 7, the positioningaccuracy of the enhanced schemes outperforms using either theSPNTV or the fusion algorithm all the time. In particular, e-SPNTV, e-fusion achieve a positioning error ≤ m with 80%and 70% availability compared to 71%, 56%, 45%, and 38%availability when SPNTV, DL-TDOA, fusion, and GNSS areused, respectively. Essentially, the e-SPNTV method allowsV2X UEs to dynamically activate and deactivate the GNSSmeasurements based on their location. This can be interpretedas another degree of freedom, in which, the selection parame-ter η is dynamically configured based on the V2X UE locationrather than using a single static value all the time. Further, thee-fusion method allows V2X UEs to overcome the inevitableGNSS performance degradation due to the tunnel losses byusing only the cellular measurements inside the tunnel.Finally, the findings of Fig. 7 are the same when theRSUs are not deployed at the tunnel edges except with slightperformance degradation due to the DL-TDOA performance degradation. In this regard, we investigate the performance ofthe most vulnerable V2X UEs in three scenarios (1) urban roadwithout a tunnel, (2) tunnel without RSUs, and (3) tunnel withRSUs using the e-Fusion and e-SPNTV methods. Fig. 8 showsthat deploying just two RSUs at the tunnel edges significantlyimproves the tail of the UE’s positioning error distribution,i.e., the positioning performance of the most susceptible V2XUEs to the high positioning errors, by more than ≈ ONCLUSION
In this paper, we investigate the potential solutions toprovide the NR V2X UEs with long-term and high-precisionpositioning in the future 3GPP releases. Specifically, wepropose a novel selection positioning method to dynamicallyswitch between the GNSS and DL-TDOA measurements basedon the locations of V2X UEs and the accuracy of the collectedmeasurements. Further, we utilize a MAP-based estimator toevaluate the performance gains of fusing the measurementsfrom different technologies in NR V2X services. The proposedhybridized methods are evaluated via extensive system-levelsimulations which demonstrate that the proposed algorithmscan achieve a positioning error ≤ ≈
76% availabilitycompared to ≈
51% and 58% availability when GNSS and DL-TDOA measurements are solely used, respectively. Further, weleverage the deployment of the RSUs to improve the tail ofthe UE’s positionig error distribution by more than ≈ EFERENCES[1] 5G Automotive Association, “C-V2X use cases: methodology, exam-ples, and service level requirements,” Tech. Rep. 5GAA-191906, Jun.2019.[2] Intel, Corporation, Ericsson, “New WID: NR positioning support,” Tech.Rep. 3GPP RP-190752, Mar. 2019.[3] LG, Electronics, FirstNet, “New SID: Study on scenarios and require-ments of in-coverage, partial coverage, and out-of-coverage positioninguse cases,” , Tech. Rep. 3GPP RP-201384, Jun. 2020.[4] R. Keating, M. Saily, J. Hulkkonen, and J. Karjalainen, “Overviewof positioning in 5G new radio,” in
Proc. 16th Int. Symp. WirelessCommuni. Syst. , Oulu, Finland, Aug. 2019, p. 320–324.[5] ESA, Mitsubishi, Electric, Corporation, u-box, “Tp on hybrid position-ing and GNSS enhancements for TR 38.855,” , Tech. Rep. 3GPP RP-1902549, Feb. 2019.[6] R. Keating, D. Yoon, T. Tao, and H. Huang, “Opportunities andchallenges for nr rat-dependent based positioning,” in
Proc. IEEE 90thVehic. Technol.Conf. , Honolulu, HI, USA, Sep. 2019, p. 1–6.[7] Technical Specification Group Radio Access Network, “Study on NRpositioning support,” Tech. Rep. 3GPP TR38.855 v16.0.0, Mar. 2020.[8] Sony, “Considerations on RAT independent and hybrid positioning forNR,” , Tech. Rep. 3GPP RP-1902190, Feb. 2019.[9] ESA, “Performance evaluation for hybrid positioning based on GNSSand NR,” , Tech. Rep. 3GPP RP-1900236, Jan. 2019.[10] Huawei, HiSilicon, “Positioning with GNSS,” , Tech. Rep. 3GPP RP-1901578, Feb. 2019.[11] Technical Specification Group Radio Access Network, “Study on LTE-based V2X services,” Tech. Rep. 3GPP TR36.885 v14.0.0, Jun. 2016.[12] 5G Automotive Association, “Positioning with GNSS,” , Tech. Rep.5GAA RP-191906, Jun. 2019.[13] F. Perez-Cruz, C. Lin, and H. Huang, “BLADE: A universal, blindlearning algorithm for ToA localization in NLOS channels,” in