Beamforming and Open RAN

Introduction

In the previous generations of wireless networks, the user throughput improvement was achieved by allocating more spectral resources for the users and densifying the deployment of the Base Stations (BSs). Although these techniques are also important from the perspective of contemporary 5G networks, they are not sufficient to meet user throughput demands by themselves. The game-changer technique, that stands as an enabler of achieving high user throughputs is named Massive Multiple-Input Multiple-Output (M-MIMO) [1]. The main idea behind the M-MIMO is to equip BSs with antenna arrays built of tens or even hundreds of elements. By proper phase shifting and amplitude scaling of the signal transmitted (or received) by each antenna spatial beams can be formulated, e.g., to increase users’ signal-to-noise ratio (SNR), and to serve multiple users within the same time-frequency resources. This procedure is named Beamforming (see Figure 1) [2].

This blog post covers the topic of Beamforming in the context of the Open Radio Access Network (O-RAN). First, we discuss the possible implementation of beamforming in 5G networks, namely Grid of Beams (GoB). Then, some challenges regarding GoB are highlighted. Finally, some ideas for utilizing O-RAN to solve them are mentioned.

Grid-of-Beams Beamforming

There are two major approaches to beamforming in wireless networks. The first option is based on the uplink pilot signal transmitted to the BS by the UE, i.e., Sounding Reference Signal (SRS). Based on this pilot signal, BS can estimate the radio channel to the UE and perform beamforming procedures, aimed at, e.g., SNR maximization, or inter-user interference suppression. However, from the perspective of network deployment, this approach is complicated, i.e., requiring SRS management procedures, and accurate radio channel estimation. Moreover,  while following this option new beams are computed very often, increasing the signal processing overhead.

Another option is to create a static set of beams, i.e., a Grid-of-Beams. In this approach, each beam is associated with a unique Synchronization Signal Block (SSB) [3]. Instead of a complicated procedure of channel estimation from uplink pilots on the BSs side, the UE measures Reference Signal Received Power (RSRP) related to SSB in the downlink, and reports it back to the BS. On the basis of such UE reports, the BS associates the user with one of the static beams. The GoB approach is much simpler in practical implementation, e.g., in terms of signal processing computational complexity, and system design. However, one should note that utilization of tens of SSBs would significantly increase signaling overhead related to reporting of Channel State Information (CSI) by the UE. The comparison between SRS and GoB-based beamforming is depicted in Figure 2. For more details, we refer the reader to [4].

Beam Mobility Management

One of the key challenges reported in the context of the GoB beamforming approach is mobility management, i.e., associating users with proper beams [5]. In real propagation environments, RSRP related to each beam would be much affected by the reflection. Thus some beams can have local maxima and minima in the spatial domain causing, e.g., ping pong effect, or beam failure. While for low mobility users (e.g., pedestrians) this would not be a significant problem, for high mobility users (e.g., cars, trains – see Figure 3) these changes in RSRP would be rapid. From this perspective there is a demand for intelligent algorithms, that would be able to predict the sequence of beams along the high-mobility users’ road to, e.g., prevent beam failures. These further open up a great opportunity to utilize location-dependent data together with Machine Learning (ML) algorithms. One example application of ML for Beam Mobility Management is to train the algorithm (e.g., Support Vector Machine) to infer the beam with the highest SNR value, on the basis of user coordinates provided as input [6].

O-RAN use cases related to beamforming

Implementation of intelligent algorithms in contemporary mobile networks requires access to the various Key Performance Indicators (KPIs), e.g., beam failure statistics, RSRPs, and location information. Moreover, proper interfaces are necessary to enable control over the network state. From this perspective, such algorithms (e.g., based on ML) can only be implemented while following the Open Radio Access Network (O-RAN) concept (for details see [7]).  

While M-MIMO is one of the key technologies for providing high throughput in 5G networks it became a hot topic within the O-RAN Alliance community.  There are several use-cases listed covering both SRS-based and GoB-based beamforming [8]:

• Grid-of-Beams Beamforming Optimization
• Beam-based Mobility Robustness Optimization
• AI/ML Based Initial Access (SS Burst set), CSI-RS, and DMRS Configuration optimization
• L1/L2 Beam Management
• AI/ML-assisted non-GoB Optimization
• MIMO DL Tx Power Optimization, MU-MIMO Pairing, and MIMO mode selection

The highlighted L1/L2 Beam Management use-case is the one that directly corresponds to the Beam Mobility Management described in the previous paragraph. The O-RAN Alliance community highlights the problem of beam selection for high-speed users, e.g., rapid blockage of a particular beam in a specific location. Their recommendation is to utilize location-dependent data like per beam RSRP to train ML algorithms aimed at, e.g., reducing beam failures or increasing SNR. 

Conclusions

The beamforming is one of the key enablers for achieving high throughputs in contemporary 5G mobile networks. Due to less complicated implementation, it is expected that BSs would transmit a fixed set of beams following the GoB. The main challenge of GoB is Beam Mobility Management, especially when considering high-mobility users. Fortunately, high mobility users usually have predictable motion patterns, e.g., railways, and streets. On the other hand, due to radio environment characteristics, i.e., multipath, RSRPs related to beams can follow unobvious spatial distributions. This specific information constitutes motivation to deploy xApp dedicated to Beam Mobility Management. Such an xApp is expected to utilize location-dependent data and ML inference to achieve Beam Mobility Management goals, e.g., beam failure prevention. 

If you are interested in this topic, stay tuned for further blog posts related to Rimedo Labs’ developments in this area.

Rimedo’s Beam Mobility Management xApp

Following the high-level architecture proposed therein, Rimedo comes up with a more in-detail concept depicted in Figure 4.

The Beam Mobility Management xApp would be deployed in the Near Real-Time RAN intelligent Controller (near-RT RIC). The xApp would host an ML model aimed at the inference of user-to-beam association based on input location information.  The ML model would be trained in non-RT RIC using the data provided through the external location information server, and KPIs obtained through the E2 interface. After each significant update, the ML model would be transferred to xApp through the A1 interface.  The xApp would offer the flexibility of Mobile Network Operator optimization goals, e.g., minimization of beam failures, and SNR maximization. 

References

[1] E. G. Larsson, O. Edfors, F. Tufvesson and T. L. Marzetta, „Massive MIMO for next generation wireless systems,” in IEEE Communications Magazine, vol. 52, no. 2, pp. 186-195, February 2014, doi: 10.1109/MCOM.2014.6736761.
[2] E., Björnson, “Adaptive Beamforming and Antenna Arrays, Commentary”, 5G Technical Insights, https://bit.ly/3cPHyJJ
[3] A. Omri, M. Shaqfeh, A. Ali and H. Alnuweiri, „Synchronization Procedure in 5G NR Systems,” in IEEE Access, vol. 7, pp. 41286-41295, 2019, doi: 10.1109/ACCESS.2019.2907970
[4] E., Björnson, “Towards 6G: Massive MIMO is a Reality—What is Next?” https://youtu.be/chmZ8cdyTMc
[5] Y. Heng et al., „Six Key Challenges for Beam Management in 5.5G and 6G Systems,” in IEEE Communications Magazine, vol. 59, no. 7, pp. 74-79, July 2021, doi: 10.1109/MCOM.001.2001184.
[6] M. Arvinte, M. Tavares and D. Samardzija, „Beam Management in 5G NR using Geolocation Side Information,” 2019 53rd Annual Conference on Information Sciences and Systems (CISS), 2019, pp. 1-6, doi: 10.1109/CISS.2019.8692820.
[7] M. Dryjanski, R. Lundberg, “The O-RAN Whitepaper; Overview, Architecture, and Traffic Steering Use Case”, 2021, https://rimedolabs.com/blog/the-o-ran-whitepaper/
[8] O-RAN Massive MIMO Use Cases Technical Report 1.0

Related Rimedo Labs Resources

• The O-RAN Whitepaper: Sign up for our newsletter to download: The O-RAN Whitepaper
• Blog post on green communications: Green Communication and Computing for 2030 – RIMEDO Labs
• O-RAN Blog posts: Other blog posts regarding the O-RAN:  1. Introduction to O-RAN: Concept and Entities, 2. O-RAN Architecture, Nodes, and Interfaces, 3. near-Real-Time RIC (RAN Intelligent Controller), 4. O-RAN Use Cases: Traffic Steering
• O-RAN website: Find out more about our O-RAN offering, services and materials O-RAN
• Webinar: A video recording from a webinar discussing O-RAN architecture, RIC internals, use cases, with a special focus on traffic steering: O-RAN Architecture and Use Cases – YouTube
• Training: Sign up for our O-RAN System Training (intelefy.com)

Author Bio