Analysis and Performance Evaluation of Conditional Handover in 5G Beamformed Systems

10/25/2019 ∙ by Umur Karabulut, et al. ∙ 0

Higher frequencies that are introduced into 5G networks cause rapid signal degradations and challenge user mobility. In recent studies, a conditional handover procedure has been adopted for 5G networks to enhance user mobility robustness. In this paper, mobility performance of the conditional handover has been analyzed for 5G mm-Wave systems with beamforming. In addition, a random access procedure is proposed that increases the chance of performing contention-free random access during the handover, which reduces signaling and interruption time. Results show that the overall failure performance improves by conditional handover scheme and the contention-free random access rate increases for proposed random access scheme.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

I Introduction

In cellular networks, demand for user data throughput has been increasing and it is foreseen that the trend will continue to increase dramatically [1]. The range of carrier frequency has been further expanded to mm-Wave frequencies in fifth generation (5G) cellular networks to meet the increasing demand of user data throughput. In addition, the number of base stations (BSs) with smaller coverage area is increased which improves frequency reuse and the total network capacity. Besides, higher carrier frequencies enable the deployment of many small-sized antennas that are used for directional signal transmission, resulting in beamforming gain.

Operating at higher carrier frequencies challenges user mobility due to steep and high diffraction loss which can lead to rapid signal degradations caused by obstacles [2]. Moreover, dense BS deployment increases the number of handovers which can cause frequent interruption of the user equipment (UE) connection, signaling overhead and latency [2].

Baseline handover (BHO) procedure that is used in Long Term Evolution (LTE) is reused for 5G networks in 3rd Generation Partnership Project (3GPP) release 15 [3, 4]. The time instant for triggering the handover in BHO is critical. This is because the signal of the source cell should be good enough to receive the handover command and the signal of the target cell should be sufficient for access. This is more visible in mm-Wave frequencies due to the rapid signal degradations and dense BS deployment.

Conditional handover (CHO) is introduced in [5] for New Radio (NR) 3GPP release 16 to increase the mobility robustness of the BHO. In CHO, the coupling between handover preparation and execution is resolved by introducing a conditional procedure, where handover is prepared early by serving cell and access towards the target cell is performed later when its radio link is sufficient. Besides, contention-free random access (CFRA) procedure is defined in [4] where target cell of handover can allocate CFRA resources for UE during the handover. Using CFRA instead of contention-based random access (CBRA) resources helps to avoid collision in random access, and consequently, higher mobility interruption and signaling overhead.

In this paper, a resource efficient random access procedure is defined such that usage of CFRA resources is increased, especially for CHO. Moreover, mobility performance of CHO is analyzed for current 3GPP and proposed random procedures, and compared against BHO.

The paper is organized as follows. The UE measurements that are used in handover are presented along with BHO and CHO in Section II. Random access procedure that is defined in 3GPP is revisited and our proposed random access procedure is presented in Section III. The simulation procedures and scenario are explained in Section IV. Simulation results are presented in Section V to show the performance of CHO and BHO in 5G mm-Wave networks for different random access procedures. The paper is concluded in Section VI.

Ii UE Measurements and Handover Models

In mobile networks, it is necessary to hand off the link of a UE between cells to sustain the user connection in the network. This handover is performed using UE received signal power measurements for serving and neighbor cells and by following a certain handover procedure. In this section, baseline handover and conditional handover procedures are reviewed along with the relevant UE measurements for mobility.

Ii-a UE Measurements in New Radio Beamforming System

A UE in the network monitors the Reference Signal Received Power (RSRP) (in dBm) at discrete time instant for beams of cell . The separation between the instants is given by ms. The physical raw RSRP measurements are inadequate for handover decisions since those measurements fluctuate over time due to fast fading and measurement errors which can cause instable handover decisions. To mitigate those channel impairments, a moving average Layer-1 (L1) filtering and infinite impulse response (IIR) Layer-3 (L3) filtering are applied by UE to RSRP measurements sequentially. Implementation of L1 filtering is not specified in 3GPP standardization and it is UE specific, i.e., it can be performed either in linear or dB domain. L1 filter output is expressed as

(1)

where is normalized by time step duration , and is the number of samples that are averaged at each L1 measurement period. For cell quality derivation of cell , set of beams having measurements above threshold is determined as

(2)

Subset consists of beams of having strongest and L1 RSRP measurement of those beams are averaged to derive L1 cell quality of cell as

(3)

Cardinality of the set is denoted by and set is adopted as in case . If is empty, is equal to stronger .

L1 cell quality is further smoothed by IIR L3 filtering and L3 cell quality output is derived by the UE as

(4)

where is the forgetting factor that controls the impact of older measurements and is the filter coefficient of IIR filter [4].

Similarly, L3 beam measurement of each beam is evaluated by L3 filtering of L1 RSRP beam measurements as

(5)

where can be configured differently from .

Herewith, L1 RSRP beam measurements , L3 cell quality measurements and L3 beam measurements of that are used during the handover and random access channel (RACH) procedure are illustrated in Figure 1.

Figure 1: Diagram of L1 and L3 UE measurements which are derived from Reference Signal Received Power (RSRP) for beams of cell .

Ii-B Baseline Handover

L3 cell quality measurements are used to assess the quality of the radio links between the UE and its serving and neighboring cells. To this end, UE reports the L3 cell quality measurements and beam measurements to its serving cell if the following condition (A3)

(6)

expires at time instant for any cell . Cell-pair specific offset can be configured differently by serving cell for each neighbor cell and time-to-trigger is the observation period of condition (6) before triggering measurement report.

After receiving L3 cell quality measurements, the serving cell sends handover request to a target cell , e.g., typically the strongest cell, along with the L3 beam measurements of target cell . Then, target cell reserves CFRA resources (preambles) for beams with the highest power based on reported . Target cell prepares the handover command including reserved CFRA resources and sends it to source cell as part of the preparation acknowledgment to serving cell. After that, serving cell sends the handover command to the UE where the command comprises the target cell configuration and CFRA preambles that are reserved by the target cell . After receiving the handover command, UE detaches from the serving cell and initiates the random access towards the target cell.

In this handover scheme, the radio link between UE and serving cell should be good enough to send the measurement report in the uplink and receive the handover command in the downlink. This is necessary but not sufficient condition for completing the handover successfully. In addition, the radio link quality between the UE and the target cell should also be sufficient so that the signaling between UE and the target cell is sustained during the RACH procedure. In a typical system level mobility simulation, the link quality of the UE is assessed by Signal-to-Interference-Noise Ratio (SINR). Herein, the link quality conditions for successful handover between serving cell and target cell are expressed as

(7a)
(7b)
(7c)

where and , are the SINR of the links between UE and the beam of target cell and serving cell , respectively, at time . is the time instant of the measurement report is sent and is the latency of handover preparation between serving and target cell. is the SINR threshold that is required for maintaining radio communication between UE and network (e.g. dB).

As shown in (7a), (7b) and (7c), the time instant for triggering the measurement report is critical for the success of handover. Delaying helps more the condition in (7b) and (7c) to be fulfilled for serving cell and target cell , respectively, at the expense of having weaker for serving cell risking the condition of (7a), and vice-verse.

Ii-C Conditional Handover

In conditional handover, the handover preparation and execution phases are de-coupled which helps to receive get the handover command safely from source cell and to access the target cell later when its radio link is sufficient.

Similar to A3 condition (6), an Add condition is defined as,

(8)

where is defined as add offset. The UE sends the measurement report to serving cell at if the Add condition is fulfilled for seconds. Then, the serving cell prepares sends handover request to target cell for the given UE. Preparation of the handover is performed as in the baseline handover where target cell reserves CFRA RACH resources for UE and sends handover command to UE via source cell. Unlike baseline handover, UE does not detach from source cell and initiate the RACH process towards the target cell when handover command is received. Instead, UE continues measuring received signals from neighboring cells and initiates the random access when the Execution condition expires at time instant , after which is defined as,

(9)

Execution condition offset is configured by serving cell and forwarded to the UE in handover command along with CFRA resources reserved by the target cell.

Smaller values lead to early preparation of the target cell and reservation of the RACH preambles which ensures that the UE sends measurement report and receives handover command, see (7a) and (7b). Besides, unlike baseline handover, lower does not lead to any early RACH attempt of the UE towards the target cell since the random access is initiated only if Execution condition is fulfilled. Higher Execution condition offset values cause the UE to perform random access late enough such that it is more likely that the is above , see(7c).

Iii RACH Procedure in New Radio Multi-Beam System

In this section, basics of random access are reviewed. Then, 3GPP RACH procedure of NR is described and the proposed RACH procedure is introduced.

Iii-a Contention-free and Contention-based Random Access

Random access is the first signaling performed by any UE for establishing the synchronization with a cell. UE initiates the random access by sending a RACH preamble to the target cell. However, it is possible that multiple UEs use the same preamble during the random access towards the same reception beam of a target cell. In this case, RACH collision occurs which is then resolved by additional signaling and delay for completing random access. This type of random access when UE selects one preamble out of set that is common for all UEs is called CBRA.

In handover, the collision risk can be avoided by assigning dedicated preambles for each UE to be used towards a prepared beam of the target cell . The network identifies the UE signal without further signaling and delay if the UE performs access to the prepared beam using the dedicated preamble. This kind of random access is called CFRA.

Iii-B Access Beam and Preamble Selection

During handover, accessing the target cell by using dedicated CFRA preamble is preferable due to less latency and signaling requirements than CBRA. Although a set of beams of the target cell with the strongest L3 beam quality measurements can be prepared with CFRA resources, measurements of those beams may vary between the preparation time instant and access time due to the de-coupling between the phases. Variation of beam measurements is more significant in conditional handover compared to baseline handover. This is because, in baseline handover, elapsed time between the preparation and access phases is given by in (7c). However, in conditional handover, this time is longer than since the UE waits until the Execution condition in (9) is fulfilled after receiving handover command. In CHO, can be much larger than .

Figure 2: Random access flow diagram. The diagram shown in black is defined in 3GPP standardization and the green block is the proposed enhancement for the random access procedure.

Due to the temporal variation of beam measurements, access beam is selected based on measurements at time instant of CHO execution. This is illustrated in Figure 2. Herein, UE selects the access beam from set of prepared beams as follows

(10)

where is the threshold that L1 RSRP beam measurements shall exceed to consider prepared beams for access. Ultimately, UE accesses prepared beam that satisfies the condition (10) and uses the corresponding CFRA preamble. If none of the measurements of beams is above the threshold , beam with the strongest L1 RSRP beam measurement is selected as

(11)

In 3GPP standardization, CBRA preambles are used if none of the L1 RSRP measurement of prepared beams is above the threshold . This has the disadvantage that the UE may make CBRA although there are CFRA resources associated with selected strongest beam. To tackle this issue, an enhancement is proposed as shown in green color in Figure 2. Herein, UE uses CFRA resources if the selected beam is prepared beam even when L1 RSRP beam measurement is below threshold . This would eventually leads to less signaling and latency during the RACH procedure.

Iv Simulation Scenario and Parameters

In this section, the investigated scenario, mobility and propagation parameters are described. These will be used to compare the different mobility performance indicators of BHO and CHO for 3GPP and proposed RACH procedures and for various random access beam threshold .

Figure 3: Madrid Grid layout is used for simulation scenario as described in METIS 2 project [6]. The scenario consists of buildings (grey), streets (black) with 200 users, open square (blue) with 40 users and pedestrian area (green) with 80 users.

In this study, the Madrid Grid layout that is described in the METIS 2 project [6] is used. The layout is given in Figure 3 and consists of buildings (grey), streets (black), open place (blue) and pedestrian area (green). There are 33 3-sector macro cells which are located on the roof tops of the buildings. The users are distributed as follows: 200 users are moving in the streets with km/h in both directions. Besides, 40 pedestrian users are walking in the open square and 80 users are walking in the pedestrian area with km/h.

The scenario parameters are defined in Table I along with the configuration of the transmit antenna panels. Beams have smaller beamwidth and higher beamforming gain to cover far region of the cell coverage area where beams with larger beamwidth and relatively smaller beamforming gain are defined to serve regions near to the cell. SINR of a link between UE and beam of cell is evaluated by the approximation given in [7] for strict resource fair scheduler.

Parameters Value
Carrier frequency GHz
System bandwidth MHz
PRB bandwidth MHz
Downlink TX power dBm/PRB
TX antenna height m
TX Antenna element pattern Table 7.3-1 in [8]
TX panel size
TX vertical antenna element spacing
TX horizontal antenna element spacing
Beam azimuth angle
Beam elevation angle
Beamforming gain model Fitting model of [9]
RX antenna height 1.5m
RX antenna element pattern isotropic
RX antenna element gain dBi
Thermal noise power dBm/PRB
Propagation loss deterministic model of [10]
Penetration loss dB
Fast fading model Abstract model of [9]
Scenario UMi-Street Canyon [8]
Network topology Madrid grid [6]
Number of cells
Total number of UEs
Number of simultaneously scheduled beams per cell
Cell-pair specific offset dB
Add offset dB
Execute offset dB
Time step size ms
L1 measurement period
Handover preparation time ms
Number of averaged samples
Table I: Simulation Parameters II

Handover Failure Model: Handover failure (HOF) is a metric that is used to evaluate the mobility performance. For both 3GPP and proposed RACH procedures, UE decides to use either CBRA or CFRA preamble as it is shown in Figure 2 and attempts to access the selected beam of target cell with selected preamble. For successful random access, it is required that the SINR of the target cell remains above the threshold , during RACH procedure. A handover failure timer ms is started when UE starts the random access and sends RACH preamble. The RACH procedure in Figure 2 is repeated until successful RACH attempt is achieved or expires. In the handover failure model, UE may succeed to access to target cell only if the exceeds the threshold . HOF is declared only if expires and UE fails to access target cell, i.e., . Once HOF is declared, UE performs re-establishment which requires additional signaling and causes latency [4].

Radio Link Failure Model: Radio link failure (RLF) is another key metric that is relevant for mobility performance. An RLF timer ms is started when SINR of serving cell falls below and RLF is declared if expires. During the timer, the UE may recover before detecting RLF if SINR exceeds second threshold which is higher than . A detailed explanation of the procedure is given in [4].

V Performance Evaluation

In this section, the proposed RACH procedure is compared against that of 3GPP for BHO and CHO. The key performance indicators (KPIs) used for comparison are explained below.

V-a KPIs

V-A1 CBRA Ratio (RCbra)

Total numbers of successful CBRA and CFRA procedures that are observed during a mobility simulation are denoted by and , respectively. The ratio of CBRA events in a simulation is formulated as

(12)

V-A2 Hof

Total number of HOFs that are observed during a simulation.

V-A3 Rlf

Total number of RLFs that are declared in the network.

Both and are normalized to number of UEs and simulation time as illustrated in the following section.

V-B Simulation Results

Mobility performance of 3GPP and proposed RACH procedure is investigated for both CHO and BHO in Figure 3. To this end, impacts of different beam access threshold values and number of prepared beam on aforementioned mobility KPIs of Section V-A are analyzed. Figure 4 and Figure 5 show the number of handover failures per UEmin with solid line on the left axis and CBRA ratio with dashed line on the right axis as a function of (in dB) for CHO and BHO, respectively. The results are shown for both proposed and 3GPP RACH procedures as well as for different number of prepared beams and .

CHO Analysis: Figure 4 shows that for the UE uses only CFRA preambles () for both proposed and 3GPP RACH procedures since UE can always select a prepared beam from set of of target cell . On the other hand, leads to worst HOF performance because received signal power of the prepared beam changes over time and prepared beam does not always remain a good candidate during the time between handover preparation and execution phases. Ultimately, the SINR of the accessed beam falls below which leads to HOF. This is more visible for since UE does not have any other options for prepared beams. Increasing from to access failure reduces to one third of its value since it increases the chance of selecting the strongest beam.

Figure 4: The number of HOFs and ratio are shown for CHO as a function of beam access threshold with RACH procedure and number of beams as parameters.

For increasing values of access threshold , RACH beam selection procedure prioritizes the L1 RSRP beam measurements and UE becomes less persistent on selecting one of the prepared beams. As a consequence, beams with higher are selected to be accessed which yields higher and less HOFs. On the other hand, for higher , UE tends to select prepared beams less frequently which results in use of CBRA preambles for random access. However, it is observed that the ratio of CBRAs is much smaller in proposed RACH procedure for higher . This is because the UE still performs CFRA if none of the prepared beams have beam measurements above threshold .

Results in Figure 4 also show that the number of HOFs of proposed and 3GPP RACH procedures reaches its lowest value at and is the same for both and . This is because the beam of the target cell with the strongest L1 RSRP measurement is selected in both RACH procedures regardless of the set of prepared beams . Hence, the selected beam of target cell with strongest measurement leads to higher SINR and in turn lower HOF. However, CBRA ratios of proposed and 3GPP RACH procedures diverge significantly at . In particular, in 3GPP RACH procedure, UE selects only CBRA preambles for random access for any value since all prepared beams have L1 measurements that are below . This is not the case for proposed RACH procedure because preamble selection still considers the prepared beams although that L1 measurement is not above .

Besides, same HOF performance of CHO is observed for both proposed and 3GPP RACH procedures since the HOF depends on the selected beam and both RACH procedures do not differ with respect to beam selection procedure as shown in Figure 2.

BHO Analysis: Figure 5 shows that HOF is not observed at BHO for any number of prepared beam and beam access threshold . This is because, compared to the CHO results in Figure 4, the time that elapses between preparation and the phases of BHO is shorter than that of CHO () and during this time the measurements of the prepared beams do not change. Consequently, UE performs access to a beam that yields sufficient at target cell .

Figure 5: The number of HOFs and ratio is shown for BHO as a function of beam access threshold with RACH procedure and number of beams as parameters.

Figure 5 also shows that the CBRA ratio of the proposed RACH procedure slightly increases for higher because the measurements of the beams do not change much between preparation and access phases which is shorter than that of CHO case. However, the CBRA ratio of the 3GPP procedure in Figure 5 gradually increases for increasing as it is observed for CHO case in Figure 4. This is also due to the fact that the 3GPP RACH procedure does not consider the prepared beams in case the L1 measurements are below the access threshold .

Failure Results: Figure 6 shows the total number of failures per UEmin as a function of beam access threshold for both CHO and BHO. As it has been shown in Figure 4 and 5 that failure rate is independent of RACH procedure, the results in Figure 6 are not distinguished for different RACH procedures. Besides, total number of failures for BHO is the same for both and .

Figure 6: The total number of failures is shown for CHO and BHO case as a function of beam access threshold with number of beams as parameter.

It is shown in Figure 6 that the overall failure performance of BHO is improved by conditional execution mechanism that is introduced by CHO. Besides, one can also state that the failures that are observed in the mobility scenario are dominated by RLF and this is improved by CHO despite the HOF increase that is observed for CHO compared to BHO, see Figure 4 and Figure 5.

Vi Conclusion

In this paper, conditional handover of 3GPP release 16 is analyzed for NR beamformed systems. Baseline and conditional handover procedures have been reviewed along with L1 and L3 UE measurements that are relevant for mobility. In addition, 3GPP random access procedure is revisited and a new random access procedure is proposed that aims to increase contention-free random access and reduce in turn signaling overhead and latency during handover. The mobility performance of conditional handover is compared against baseline handover. Simulation results have shown that the number of fall-backs to contention based random access has reduced significantly when proposed random access procedure is used.

Moreover, results have revealed that the baseline handover procedure causes less handover failures than conditional handover. However, the total number of failures for conditional handover is less than that of baseline handover due to the de-coupled handover preparation and execution phases, providing mobility robustness.

References

  • [1] Cisco, “Cisco visual networking index: Global mobile data traffic forecast update 2017–2022,” Cisco, Tech. Rep., Feb 2019, White Paper c11738429.
  • [2] M. Tayyab, X. Gelabert, and R. Jäntti, “A survey on handover management: From lte to nr,” IEEE Access, vol. 7, pp. 118 907–118 930, 2019.
  • [3] 3GPP, “NR overall description stage-2,” 3rd Generation Partnership Project (3GPP), Tech. Rep. 38.300, Sep 2019, V15.7.0.
  • [4] ——, “NR radio resource control protocol specification,” 3rd Generation Partnership Project (3GPP), Tech. Rep. 38.331, Jun 2019, V15.6.0.
  • [5] ——, “Conditional handover – basic aspects and feasibility in Rel-15,” 3rd Generation Partnership Project (3GPP), Tech. Rep. TSG-RAN WG2 NR Adhoc 2, Jun 2017, R2-1706489.
  • [6] P. Agyapong et al., “Simulation guidelines,” Mobile and wireless communications Enablers for the Twentytwenty Information Society (METIS 2), Tech. Rep., 2013, Deliverable ICT-317669-METIS/D6.1.
  • [7] A. Ali et al., “System model for average downlink sinr in 5g multi-beam networks,” in 2019 IEEE 30th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), September 2019, pp. 1–6.
  • [8] 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,” 3rd Generation Partnership Project (3GPP), Tech. Rep. 38.901, Jun 2018, V15.0.0.
  • [9] U. Karabulut, A. Awada, I. Viering, A. N. Barreto, and G. P. Fettweis, “Low complexity channel model for mobility investigations in 5G networks,” https://arxiv.org/abs/1910.10438, 2019.
  • [10] A. Awada, A. Lobinger, A. Enqvist, A. Talukdar, and I. Viering, “A simplified deterministic channel model for user mobility investigations in 5g networks,” in IEEE International Conference on Communications (ICC), May 2017, pp. 1–7.