Indoor positioning is currently recognized as one of the important features in emergency, commercial and industrial applications. The 5G network enhances mobility, exibility, reliability, and security to new higher levels which greatly bene t the IoT and industrial applications. Industrial IoT (IIoT) usecases are characterized by ambitious system requirements for positioning accuracy in many verticals. For example, on the factory oor, it is important to locate assets and moving objects such as forklifts. The deployment design for di erent IIoT environments has a signi cant impact on the positioning performance in terms of both accuracy and availability of the service. Indoor factory (InF) and indoor open o ce (IOO) are two available and standardized Third Generation Partnership Project (3GPP) scenarios for evaluation of indoor channel models and positioning performance in IIoT use cases. This paper aims to evaluate the positioning performance in terms of accuracy and availability while considering di erent deployment strategies. Our simulation-based evaluation shows that deployment plays a vital role when it comes to achieving high accuracy positioning performance. It is for example favorable to deploy the 5G Transmission and Reception Points (TRPs) on the walls of the factory halls than deploying them attached to the ceiling.
Localization in cellular networks is primarily used to locate a user equipment (UE) in
outdooronly scenarios, often by exploiting global navigation satellite system (GNSS) that can guarantee
meter-level accuracy. In the recent years, mainly since the third generation partnership project
(3GPP) Release 13 [
In the 5G positioning study in 3GPP Release 16 [
One important aspect in the outcome of any cellular positioning method is the deployment
of the nodes which perform the transmission and reception of the signals. It is well-known that
the number of nodes and how they are distributed within the serving area has a great impact
on the positioning accuracy. Due to this dependency, there is always a trade-o between how
accurate the position estimation can be and how complex and costly the positioning deployment
is [
There is already an extensive research on the topic of indoor positioning exploiting different technologies. However, most of these studies aim to explore and study the accuracy and the potential of each indoor positioning solution assuming a xed deployment which is typically optimized to provide decent communication services. Analyzing the impact of di erent deployment strategies on positioning performance therefore remains unexplored.
In this paper, our aim is to analyze di erent aspects of TRP deployment strategy and
understand its impact on 5G indoor positioning performance. The results are then analyzed for
the purpose of understanding how positioning accuracy and availability relates to di erent
deployments. This paper serves as a comprehensive summary of a well-studied thesis work
[
To provide a good overview of the conducted study, this paper is organized in six sections. Section 2.1 introduces downlink time di erence of arrival (DL-TDOA) method which is the considered positioning technique, as well as the two analysis methods Cramér-Rao lower Bbund (CRLB) and geometric dilution of precision (GDOP). Section 3 provides introduction to the indoor open o ce (IOO) and indoor factory (InF) scenarios as de ned by 3GPP. Moving on, simulation results and performance evaluations are presented in sections 4 and 4.4 while Section 5 draws some conclusive remarks.
In this section, we rst de ne the DL-TDOA method and further explain how we assess a lower bound on the positioning accuracy. Finally, we explain how we evaluate the impact of deployment geometry independent and decoupled from other positioning errors.
The 5G DL-TDOA is a positioning method, in which the UE measures the downlink time of arrival (TOA) of positioning reference signals (PRSs) from di erent TRPs, and reports the time di erence of arrival measurements for these TRPs to the location server in relation to the timing of a reference TRP. The positioning estimation is then done by performing multilateration based on the UE reported timing measurements.
Once the timing measurements are obtained, they are translated to distance observations by multiplying it with the speed of light. A general measurement equation at time t has the form, yt = h(✓ t) + et, where yt is the measurement, h(✓ t) is a nonlinear measurement model and et is noise. All mentioned variables are vectors. The variable ✓ t = [xt yt zt]T is the 3D position of the UE. In general, the function h(✓ t) will implicitly depend on the known positions of the N TRPs, pi = [xi yi zi]T , i 2 { 1, 2, . . . , N }.
A TOA observation based on TRP i 2 { 1, 2, . . . , N } in an asynchronous network can be expressed by,
i yTOA = |✓
pi| + i + ei,
where i is the unknown clock bias between the UE and TRP i [
It is often useful to state a lower bound on the performance of any unbiased estimator. The
CRLB which expresses a lower bound on the variance of unbiased estimators of a deterministic
( xed, though unknown) parameter, stating that the variance of any such estimator is at least as
high as the inverse of the Fisher information, serves that purpose. In order to compute the CRLB
one needs to rst determine the Fisher information matrix (FIM),I(✓ ). When the measurement
noise is additive white Gaussian noise (AWGN), the FIM becomes [
i yTOA = |✓ pj| | ✓ pi| + ej ei, where i denotes reference TRP and j denotes another TRP for which timing measurements are performed by the UE.
I(✓ ) = HT (✓ )R 1H(✓ ),
Cov(ˆ✓) I
1(✓ ).
where H(✓ ) = r ✓ h(✓ ) and R is the measurement noise covariance matrix. The information is
additive for independent observations [
RMSE = pE [(x xˆ)2 + (y yˆ)2 + (z
zˆ)2] ptr(I 1(✓ )).
RMSE = pE [(x q xˆ)2 + (y
yˆ)2]
I1,11(✓ ) + I2,12(✓ )
RMSE of an estimator is obtained by taking the square root of the trace of CRLB:
with I1,11(✓ ) and I2,12(✓ ) being the diagonal elements of I 1 corresponding to the x and y
directions [
We seek to decouple the e ect of the TRP deployment geometry on the positioning error
from the e ects of other measurement errors. For that we use the (GDOP), which intends to
state how errors in the measurement will a ect the nal state estimation1[
For more details regarding GDOP, see e.g. [
In this study two representative indoor scenarios, IOO and InF, de ned in 3GPP are considered
[
The IOO scenario is de ned as a deployment area of dimensions 120 m⇥ 50 m ⇥ 3 m designed
to capture typical indoor environments such as shopping malls and o ces. The TRPs in this
type of environment are typically mounted on the ceiling at a height of 3 m [
The InF scenario represents factory halls of varying sizes and clutter densities. The area is 120 m ⇥ 60 m with a ceiling height of 5–25 m where the TRPs can be mounted.
The InF scenario has ve di erent variants. Out of the ve, two variants named indoor
factory-sparse high (InF-SH) and indoor factory-dense high (InF-DH) are chosen for this study
as they depict realistic InF scenarios with sparse and dense clutter densities. The sparse clutter
option speci es an industrial factory oor whose< 40% of the area is covered by machines,
storage shelves, assembly lines, etc. Likewise the dense clutter option speci es an industrial
factory oor with 40% of the area covered by machines, storage shelves, assembly lines, etc.
Descriptions of both InF scenarios are presented in Table 1.
3.3. Deployment Strategies
3GPP has de ned standard TRP deployments for both IOO and InF. The proposed deployments
are mainly motivated to meet the quality of service (QoS) to support communication
requirements of IIoT use cases. In IOO scenario 12 TRPs are deployed in two rows and in InF scenario
18 TRPs are deployed in three rows. In both scenarios an inter site distance (ISD) of 20m is
maintained in both x and y axes [
• Mixed deployment:A mix of the standard and edge deployment strategies. The IOO scenario is analyzed with the deployment strategies illustrated in Figures 1(a)–(c) and the InF scenario with the deployment strategies in Figures 1(a), (b) and (d). Notice that slightly di erent mixed deployments are used for IOO and InF (compare Figures1(c) and (d)). The triangles in the gures represent the 5G TRPs and the red lines show the deployment area boundaries. Apart from these, in Section 4.3 we analyze the e ect of TRP densi cation in positioning accuracy in IOO scenario. (a) Standard deployment (IOO and InF). (b) Edge deployment (IOO and InF).
In this section we show simulation results taking into account realistic channel models that depict
typical propagation environments in IOO and InF scenarios. An internal Ericsson simulation
tool is used which applies the channel model and parameters agreed by 3GPP in [
In Fig. 2(a) and 2(b), contour plots of the GDOP for the standard and edge deployments in IOO are presented. The GDOP is in general lower and more homogeneous in the edge deployment in comparison to the GDOP when the deployment is based on standard layout. The standard deployment procures worse GDOP at the corners and along the short sides of the deployment area. The contour plots indicate that the edge deployment is more favorable in terms of receiving high positioning accuracy compared to the standard deployment as the dilution of positioning accuracy due to the geometry of the TRP deployment is lower.
Fig. 2(c) and 2(d) show the CRLB (Equation (7)) of the UE position estimates obtained with DL-TDOA. What can be noted is that the CRLB is larger for the standard deployment than for the edge deployment. There are also much larger variations in CRLB for the standard deployment compared to the edge deployment. In the standard deployment, the CRLB range is from 0.49m to 3.94m while in the edge deployment the range is from 0.47m to 0.95m. These observations show that the measurement error is not the only factor that limits the achievable UE localization accuracy. The e ect of the geometry of the TRP deployment is also of paramount interest.
NLOS conditions between the TRP i 2 { 1, . . . , N } and the UE adds a positive o set to Equation (2) since the signal has traveled longer than the euclidean distance |✓ pi|. The o set can result in a biased position estimate.
Fig. 3 visualizes the cumulative distribution functions (CDFs) of the positioning error for the standard, edge and mixed deployments in IOO when either all measurements or measurements from only LOS TRPs are used. In this gure we can observe that the edge and the mixed deployments yield similar positioning accuracy and the achievable accuracy is higher than when the TRP geometry is based on standard layout. This result is in line with our theoretical analysis based on GDOP and CRLB. Moreover, the results indicate only small di erences in positioning accuracy between either using measurements from all TRPs or using the measurements only corresponding to the TRPs that are in LOS condition with the UE. This can be understood by considering the LOS statistics reported in Fig. 4. The histograms show the distribution of the number of LOS TRPs in standard and edge deployment scenarios. It is worth noting that in both scenarios there are very few UEs with less than four LOS links, which is the smallest number of links required to obtain a position estimate using DL-TDOA, and therefore the positioning error distribution looks similar when the UE position is estimated using measurements from all TRPs and using measurements from only LOS TRPs. Moreover, more UEs can maintain a LOS condition when the TRP deployment follows edge layout.
250 200 s frEU150 o e b um100 N 50 250 200 sE150 U fr o e b um100 N 50 02 3 4 5 6 7 8 9 10 11 12 13
Number of LOS links (a) Standard deployment.
Fig. 5 shows the CDFs of the positioning error for the standard, edge and mixed deployments in the InF-SH scenario. It can be observed that in general the achievable positioning accuracy is higher than in IOO scenario and the edge deployment gives better accuracy than the standard layout deployment of the TRPs. The higher accuracy positioning in InF-SH, in comparison to the IOO, is due to higher probability of TRPs being in LOS with the UE. Fig.6a and Fig.6b show the LOS statistics for InF-SH standard and InF-SH edge deployments. The LOS statistics for InF-DH standard and InF-DH edge deployments are reported in Fig. 6c and Fig.6d. With majority of UEs having no LOS link with any of the deployed TRPs InF-DH, where more than 40% of the area is covered by clutters, is a challenging environment for precise UE positioning. In the sub-section to follow we address the problem of enhancing positioning accuracy in InFDH scenario by densi cation of TRP. Moreover, the e ect of TRP densi cation on achievable positioning accuracy in IOO scenario is also evaluated.
The e ect of TRP densi cation on achievable positioning accuracy in IOO scenario (both standard and edge) is shown in Fig.7. It has been observed that the positioning accuracy improves when the number of TRP is increased. When the number of TRP, following the standard deployment layout, is increased beyond 36 the achievable positioning accuracy tends to saturate. Similar observation can also be made when the number of TRP is increased following the edge deployment layout. These observations validate that in either of the deployment strategies the improvement in positioning accuracy is limited by the number of TRP that can be deployed in a scenario like IOO.
Densi cation of a given deployment can alternatively be done by successive addition of TRPs in areas where the accuracy is bad. Fig. 8(a) shows the location of UEs with the largest positioning errors in InF-DH scenario when the deployment is based on mixed layout. It has been observed that the successive addition of TRPs helps improving the achievable positioning accuracy. In particular the 90% positioning accuracy improves from 13.11m to 12.13m when one additional TRP is deployed in the area where the positioning accuracy is bad. Deploying two more TRPs to the problematic area helps improving the positioning accuracy from 13.11m to 11.80m. Furthermore, the achievable positioning accuracy is 11.55m when three additional TRPs are deployed in the problematic area as shown in Fig. 8(b)–(d). The CDF curves showing the positioning error in the InF-DH scenario when the deployment strategy follows a mixed layout is shown in Fig. 9 where it can be observed that the overall positioning accuracy only improves slightly when TRPs are added.
(a) Standard deployment in InF-SH.
0 2 4 6 8
Number of LOS links (b) Edge deployment in InF-SH.
10 700 600 0 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Number of LOS links (c) Standard deployment in InF-DH.
0 -0.5 0 0.5 1 1.5 2 2.5
Number of LOS links (d) Edge deployment in InF-DH.
3 3.5
In this paper we presented our study on achievable positioning accuracy in IOO and variety of InF scenarios where the focus is to verify the e ect of deployment strategies and TRP densi cation on DL-TDOA positioning technique. In IOO scenario, we observed e ects of various strength 2.5 ]m 2 [rrr o e ign1.5 n o iit s oP 1 0.5 coming from the number of TRPs in the deployment, geometry of the deployment, and LOS conditions that a ect the achievable positioning accuracy. The observations can be summarized as: • An increasing number of TRPs proves to always improve the positioning accuracy and also the availability. • The deployment geometry is of high importance and must be considered as one of the important factors for high accuracy UE localization. • Better LOS conditions are favorable from a positioning perspective and improve the positioning accuracy.
The study of the two InF scenarios highlighted the di culty in achieving high accuracy and availability indoor positioning in a densely cluttered deployment area in comparison to a sparsely cluttered deployment area. The vital factor was an immense lack of the number of LOS links which proved that decent LOS conditions are necessary for accurate positioning. The most prominent di erence between IOO and InF is that the clutter in InF, especially InF-DH, leads to a great loss in LOS links. This makes the trade-o between favorable geometry and optimizing the LOS links more di cult in InF-DH than in IOO, where the geometry has the biggest impact on the results. Table 3 and Table 4 summarizes the achievable positioning accuracy given that the deployment is adjusted based on studied strategies in the studied scenarios.
In this paper, we investigated di erent aspects of deployment and its impact on 5G indoor positioning. While the standard deployment designs proposed by 3GPP always considers the TRPs to be in the ceiling distributed evenly inside the indoor hall, our studies show that for improved positioning performance, it is more favorable to mount the TRPs evenly on the wall at the edges of the deployment area. In a more reasonable approach, in which both positioning and communication performance requirements are considered, the strategy would be to have a mixed deployment in which most of the TRPs are on the walls and few distributed on the ceiling if the area of deployment is speci cally cluttered due to the presence of machines and other objects in a factory oor. A nal interesting nding is that if one has to choose between two deployments, one promoting deployment geometry and the other promoting the number of LOS links, the choice should fall on the rst option.
This work has been supported in parts by Ericsson Research AB and by the European Union’s Horizon 2020 research and innovation programme under grant agreement No.871249 (Research and Innovation Action), LOCalization and analytics on-demand embedded in the 5G ecosystem for Ubiquitous vertical applicationS (LOCUS).