Beamforming Optimization – Use Cases

The Massive Multiple-Input Multiple-Output (M-MIMO) [1] has progressed from a theoretical innovation to a practical implementation in 5G networks. As operators increasingly rely on antenna arrays to meet strict demands, such as better spectral efficiency, coverage, or throughput, the problem is no longer whether the M-MIMO works, but how to control and optimize it efficiently. One approach to realize M-MIMO systems is the Grid of Beams (GoB) concept (see our earlier post). The GoB relies on a predefined set of beams that are launched in fixed directions. User Equipments (UEs) report the Reference Signal Received Power (RSRP) associated with each beam, and the Base Stations (BSs) choose the one to connect the UE to (typically the strongest one). This provides a way to use the spatial capability of M-MIMO, through beamforming optimization, without advanced channel estimation or management of uplink pilots at the BS.

However, the main disadvantage of GoB is that the fixed set of beams may not track the dynamics of a real network, like users’ mobility. In some cases, this can lead to Quality of Service (QoS) degradation, e.g., poor coverage. In this blog, we explore how to address this challenge by examining the concept of Beamforming Optimization (BO) in the context of its applicability. We present four use cases, with different approaches to the GoB, ranging from static to dynamic beam adaptation.

Beamforming Optimization Use Cases

The GoB optimization can be realized in different time scales, ranging from static BO during deployment, to once-per-week or per-month adjustments, to dynamic BO based on location information, e.g., adapted every minute. Moreover, optimization goals can vary from coverage optimization or reduction of inter-beam ping-pongs to improvement of UE throughput. To illustrate this diversity, we identified four BO use cases.

Static/Default GoB Configuration

The first case is the Static GoB optimization, as shown in Fig. 1, where the default GoB set during deployment is adjusted to maximize the coverage. This scenario is closely related to network planning. Therefore, using advanced simulators, e.g., based on Ray-Tracing tools, one can determine the most beneficial GoB configuration. The key performance metrics considered in this use case, aiming at coverage maximization, are RSRP and Signal-to-Interference plus Noise Ratio (SINR). The Static GoB Configuration is sufficient for environments with predictable coverage needs, such as rural sectors or areas with little variation in user distribution.

Event-Based/Semi-Static GoB Configuration

While the fully static approach to GoB optimization may be sufficient for serving day-to-day mobile traffic, it can suffer from poor performance under unnatural network loads. This can be related to, e.g., social events, a concert, or a football match. We refer to this use case as event-based GoB optimization (see Fig. 2). By considering the nature of sport or social events, which are typically planned, this introduces semi-static GoB reconfiguration, where the new GoB is determined based on ray-tracing simulations to deal with the expected network traffic increase of a specific place.  With that additional information, the static GoB configuration can be changed according to the one optimized for a planned event. After the event ends, the GoB configuration changes can be reverted to the default one. This occasional adjustment of GoB can significantly improve the QoS of a user who takes part in the event. Metrics to evaluate performance are similar to the static GoB optimization, i.e., RSRP and SINR. However, considering additional information about data volume, latency, or cell loads from past events can be beneficial.

Slow Dynamic GoB Configuration

Event-based GoB Configuration is a more dynamic approach to the BO optimization, considering the network conditions. One step further is to collect network data and observe the long-term daily traffic patterns to adapt the GoB configuration to the actual throughput demand.  This enables a closed-loop optimization that continuously reacts to changes in the network state. We refer to this use case as Slow Dynamic, depicted in Figure 3. It approaches the situation with periodic Key Performance Indicator (KPI)-driven optimization, where the system regularly evaluates PMs and adjusts the GoB orientation to maximize a chosen objective, e.g., maximizing UE’s overall throughput. The PMs to be considered in this use case are the distribution in time of RSRP and SINR. The timescale of the adjustment depends on long-scale dependencies in the network, e.g., morning-afternoon or weekday-weekend, when the variations are significant. To capture the dependencies, the optimization algorithm needs to analyze the UEs’ throughput at each time of day or day of the week. When such variations are noticed, the regular adjustment of the GoB may be done. To this end, location information can play an important role. Even having information about the general routes of UEs over different times of the day, GoB can be configured to point in those directions.

Fast Dynamic GoB Configuration

Finally, the most dynamic scenario, where the beams are adjusted regularly at relatively small intervals, is presented in Fig. 4, i.e, fast dynamic GoB configuration. In this case, the frequency of decisions on GoB configuration depends on, e.g., hardware limitations for O-RU, or interface limitations in terms of signalling. Typically, it is 5-15 minute intervals in practical deployments. For this purpose, the network needs the UE’s location information provided, e.g., from external data sources or predictions generated by a Network Digital Twin. This can be augmented by SINR, or RSRP measurements, in time distribution, as well as radio resource utilization per beam. Beam orientation can be adapted proactively by anticipating upcoming fluctuations in traffic distribution or users’ movements. Compared to the previous use cases, Fast Dynamic GoB Configuration represents the most advanced form of GoB optimization. It is a continuously adjusting, intelligence-driven approach that reflects both the present and future network conditions, potentially creating an opportunity to benefit from utilizing advanced Artificial Intelligence / Machine Learning (AI/ML) algorithms.

Conclusions

In this blog post, we covered a practical aspect related to Beamforming Optimization and its implementation in mobile networks. In detail, the BO targeting GoB can be applied to the 4 proposed use cases:

• Static – related to the network planning, and intended to be optimized for the average day-to-day traffic;
• Event-based/semi-static – introducing adjustment of the GoB to the planned sport or social event;
• Closed-loop slow dynamic – extending the previous use cases to track network patterns associated with a day (morning, afternoon) and adjust GoB accordingly;
• Closed-loop fast dynamic – with GoB configuration being adjusted to the current UE mobility.

While static Beamforming Optimization requires the least complicated metrics, such as RSRP and SINR, the performance gains can be limited. On the other hand, the fast dynamic BO can potentially provide high throughput improvement, but requires frequent PM reporting and the UE’s updated location information.

Rimedo Labs has developed a Beamforming Optimization rApp (BO-rApp) that addresses the use cases discussed herein – see the details in the next blog post!

References

[1] What is Massive MIMO? | Wireless Future Blog

[2] M. Hoffmann „Beamforming and Open RAN”,  Rimedo Labs Blog, [Online], Available: https://rimedolabs.com/blog/beamforming-and-open-ran/

[3] M. Pakuła „Digital Twin – What Is It and How Can It Affect Future Networks” Rimedo Labs Blog, [Online], Available: https://rimedolabs.com/blog/digital-twin-what-is-it-and-how-can-it-affect-future-networks/

Acknowledgement

The author would like to thank Marcin Hoffmann for the valuable comments provided during the writing process.

Author Bio