MmWave Codebook Selection in Rapidly-Varying Channels via Multinomial Thompson Sampling

Abstract

Millimeter-wave (mmWave) communications, using directional beams, is a key enabler for high-throughput mobile ad hoc networks. These directional beams are organized into multiple codebooks according to beam resolution, with each codebook consisting of a set of equal-width beams that cover the whole angular space. The codebook with narrow beams delivers high throughput, at the expense of scanning time. Therefore overall throughput maximization is achieved by selecting a mmWave codebook that balances between beamwidth (beamforming gain) and beam alignment overhead. Further, these codebooks have some potential natural structures such as the non-decreasing instantaneous rate or the unimodal throughput as one traverses from the codebook with wide beams to the one with narrow beams. We study the codebook selection problem through a multi-armed bandit (MAB) formulation in mmWave networks with rapidly-varying channels. We develop multiple novel Thompson Sampling-based algorithms for our setting given different codebook structures with theoretical guarantees on regret. We further collect real-world (60 GHz) measurements with 12-antenna phased arrays, and show the performance benefits of our approaches in an IEEE 802.11ad/ay emulation setting. CCS CONCEPTS• Networks → Mobile networks; • Computing methodologies → Machine learning algorithms; • Mathematics of computing → Bayesian computation.

Multi-armed Bandits for Link Configuration in Millimeter-wave Networks

“…For example, wide beams benefit devices that are moving quickly (more resilient to beam pointing error) and those that are closer to the base station (do not need the array gain derived from searching a larger codebook). Any inappropriate choice could potentially degrade the overall system throughput [2]. In this situation, the sequential decisionmaking problem is for the system to adapt the beamwidth and therefore the codebook size over time.…”

“…Each beam set consists of beams sharing the same width and together covering cover the entire spatial area. It is worth pointing out that the generation of beams of different resolutions could be achieved by beamforming optimization or antenna on-off techniques [2]. The optimal beamwidth is determined by both user channel conditions and beam sweeping period.…”

“…Many of these algorithms involve successive operations of adaptation and reconfiguration. For example, beamforming in a millimeter wave network might involve frequent beam training (finding the best beam in a beam-training codebook [1]) and a less-frequent reconfiguration (finding the best beam codebook size among several possible codebooks [2]). Solving the reconfiguration problem is challenging as the optimal solution changes over time due to dynamics in the environment and mobility in the network, and it is unrealistic to find the optimal solution at each configuration point due to large training overhead.…”

“…More recently, MAB has been applied to solve problems in wireless communications. Relevant examples include learning and tracking a rate and a MIMO mode pair that provides highest data rate in 802.11 systems [6], finding an optimal set of beams for beam training in mmWave networks [7], [8], and selecting the optimal beamwidth for mmWave beam training in our prior work [2]. Prior work has showed how MAB can improve cumulative system throughput compared to reasonable non-sequential solutions.…”

Reinforcement Learning-based Joint User Scheduling and Link Configuration in Millimeter-wave Networks

“…As a result, selecting an appropriate codebook that resolves this tension among beamforming gain, training overhead, and link robustness is critical. In particular, a dynamic approach is encouraging since the optimal codebook is generally user-dependent and scenario-dependent [15], [16].…”