Return to Session C6 Next Abstract

Session C6: Terrestrial Signals of Opportunity-Based Navigation Systems

5G Direct Position Estimation for Precise Localization in Dense Urban Area
Sijia Li, Sergio Vicenzo, and Bing Xu, The Hong Kong Polytechnic University

Objectives
Precise positioning in harsh environments is a challenging problem due to severe signal blockage and multipath effects. The traditional Global Navigation Satellite System (GNSS) positioning performance degrades substantially in such environments. The utilization of non-GNSS radio signals for positioning, navigation, and timing (PNT) has received significant attention over the past few years, and it has been considered as a viable supplement when line-of-sight (LoS) GNSS observables are scarce [1]. The fifth generation (5G) cellular signals are of particular interest for both academia and industry due to their high signal bandwidth, desirable geometric diversity, and considerable distribution density. However, a significant challenge in using 5G signals is their susceptibility to non-line-of-sight (NLoS) propagation effect, which is particularly common in urban street canyons. The resulting NLoS bias leads to an inevitable timing delay error in the ranging estimation, an outcome causing an impact on the positioning accuracy. In addition, traditional 5G positioning methods, such as round-trip time (RTT) and time-difference of arrival (TDoA), require stringent synchronization either between the user equipment (UE) and the base station (BS) or among all BSs [2]. Therefore, there is an urgent need for research into alternative algorithms that are less vulnerable to the effects of NLoS and imperfect synchronization in 5G positioning.
The direct position estimation (DPE) has been introduced as a more robust algorithm in terms of achieving precise positioning in harsh propagation scenarios, which utilizes an idea of solving the receiver position, velocity, and timing (PVT) directly via maximum likelihood estimation (MLE) in the navigation domain [3-6]. Currently, 5G direct positioning has remained unexplored. The challenge for 5G DPE lies in the fact that the channel model for 5G signal propagation in urban scenarios is much more complex than that for GNSS reception. The study on channel model for cellular signals provided by 3GPP [7] indicates three major characteristics of 5G signals:
1. Large scale multipath exists in both LoS and NLoS channels.
2. For LoS channel, it also consists multipath, but the first arriving path carries the strongest power, which is outstandingly higher than the summation of all power separated on other multipath components.
3. For NLoS channel, there is no multipath component occupying a significantly stronger power than other reflected paths, and it still has a first arriving path but has a relatively low power.
The unique cellular channel characteristic implies that the 5G DPE method may be more capable of correcting NLoS errors, as it directly sums all the first arriving path components from each contributed BS. To verify this idea, this study presents a necessary comparable analysis regarding both the DPE and the TDoA positioning performance for 5G localization in dense urban scenarios, via large scale Monte-Carlo system-level simulations. Simulation results reveal that the 5G DPE method is more competent in solving the NLoS problems in dense urban scenarios given imperfect synchronization, outperforming the traditional TDoA method at a great magnitude.
Key innovative steps:
The simulation adopts a realistic cellular distribution in a densely populated urban scenario in Tsim Sha Tsui, Hong Kong, where a total of 26 BSs are distributed within a 25000m2 squared region [8]. The localization scenario in general follows the urban macro (UMa) use case defined by the 3rd Generation Partnership Project (3GPP), where the BS is assumed to be located on the rooftop of buildings with an average height of 25m and a constant transmitting power of 250mW. The LoS probability for each BS in every simulation epoch is modelled and calculated in terms of the geometry distance between the BS and UE. Having determined the LoS condition of the propagation link, the path loss and the tapped delay line (TDL) wireless channel are stochastically generated according to the parameters given in [7]. The coarse downlink and uplink synchronization are assumed to be finished, since the downlink positioning reference signal (PRS) is utilized, which only broadcasts when the Radio Resource Control (RRC) states is connected (RRC_CONNECTED). The synchronization error either between the UE and the BS or among all BSs thus can be randomly generated with the truncated Gaussian distribution, having a standard deviation of 5ns [9]. The timing offset caused by the synchronization error is therefore modelled as a constant phase shift, linearly applying to the orthogonal frequency division multiplexing (OFDM) symbols in the frequency domain of the received signal. At the receiver side, a divide-and-conquer based DPE algorithm is deployed, which aims at generating a correlogram by correlating the received signal from a single BS with the local replica generated at each candidate point around the actual UE [4]. Instead of repeatedly calculating correlation for each candidate point, the DPE algorithm is improved by performing the correlation only once, with the local replica delayed from zero to the maximum inter-site distance (ISD) of the examined region. The pre-calculated correlation values are stored as a reference table, recording the pre-correlations corresponding with each code delay [4], [10], [11]. The final correlogram to be examined is updated by summing all BSs, and the peak point of the correlogram yields the estimated location by the DPE. The localization achieved by the traditional TDoA algorithm is derived for each simulation, for the sake of performance comparison. A total number of 5000 epochs are simulated via Monte-Carlo simulations to obtain the statistical validity for the results.
Preliminary results
The PRS occupying full 100MHz system bandwidth is used for the simulation. The subcarrier spacing (SCS) of 30KHz is used, and the baseband signal is upconverted to 4.5GHz for downlink transmission. The LoS propagation cases only occupy 17% of the total simulations, while 87% of the cases are NLoS channels, which implies the results are derived through an NLoS-dominant propagation scenario. The LoS channel is modelled as 3GPP TDL-D channel, with the first LoS path following a Ricean and the other 13 multipaths following a Rayleigh distribution. The NLoS channel is modelled as 3GPP TDL-C channel, with a total of 24 multipaths followed by a Rayleigh distribution. We use the cumulative distribution function of the horizontal error as metrics to analyze the positioning performance of the 5G DPE and TDoA methods. We observe that the DPE method manifests precise positioning performance, with a horizontal accuracy below 0.92m in 90% of cases. However, the performance of the traditional TDoA method is not very desirable due to the existence of synchronization bias and the NLoS-dominant channel condition, with a mean horizontal positioning error as large as 43.04m. The preliminary results indicate that the DPE algorithm is expected to be more appropriate for NLoS-dominant dense urban use cases, considering imperfect synchronization, in terms of 5G localization.
Conclusion
In conclusion, this abstract presents a system-level simulation of 5G DPE, underpinned by the analysis of the characteristics of the cellular propagation channel. Simulation results show that a sub-meter positioning can be achieved via 5G DPE in 90% NLoS-dominant dense urban scenarios. Although the performance relies on the contribution of a large number of terrestrial BSs, future work will be focused on the possible integration with GNSS signals, with limited 5G signals, to achieve precise 5G-GNSS DPE while decreasing the computational burden of the network.

Return to Session C6 Next Abstract