Beam Scheduling in Millimeter Wave Networks Using the Whittle Index
Mandar R. Nalavade, Ravindra S. Tomar, Gaurav S. Kasbekar · 2025
Read Beam Scheduling in Millimeter Wave Networks Using the Whittle Index by Mandar R. Nalavade, Ravindra S. Tomar, Gaurav S. Kasbekar (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-18133.txt.
Introduction
Mandar Nalavade, Ravindra Tomar, and Gaurav Kasbekar tackle beam scheduling for fast wireless networks in a 2025 paper. Millimeter wave base stations send data in narrow beams. A station can form only a few beams at once, fewer than its user count. Packets wait in queues, and waiting carries a cost.
Each slot, the station must choose which users get a beam. A sent packet lands only when the channel to that user is good. The authors frame the choice as a known hard problem type. They prove a scoring rule, the Whittle index, applies to each user here. Their policy serves the users with the lowest scores first. Long simulations show the policy beats rival rules by a wide margin.
Catalog record
| CREATOR | Mandar R. Nalavade, Ravindra S. Tomar, Gaurav S. Kasbekar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-18133
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Beam Scheduling in Millimeter Wave Networks Using the Whittle Index Mandar R. Nalavade, Ravindra S. Tomar, Gaurav S. Kasbekar Abstract We address the problem of beam scheduling for downlink transmissions in a single-cell millimeter wave (mmWave) network. The cell contains a mmWave base station (mBS) and its associated users. At the end of each time slot, a packet arrives into the queue of a user at the mBS with a certain probability. A holding cost is incurred for the packets stored in a user’s queue at the mBS in every time slot. The number of simultaneous beams that the mBS can form to different users is less than the number of associated users. Also, a cost is incurred whenever a beam is formed from the mBS to a user. In a given time slot, a packet transmitted from the mBS to a user that has been assigned a beam is successfully received (respectively, not received) if the channel quality between the mBS and the user is good (respectively, bad). In every time slot, the mBS needs to assign the available beams to a subset of the users, in order to minimize the long-run expected average cost. This problem can be modeled as a restless multi-armed bandit problem, which is provably hard to solve. We prove the Whittle indexability of the above beam scheduling problem and propose a strategy to compute the Whittle index of each user. In each time slot, our proposed beam scheduling policy assigns beams to the users with the smallest Whittle indices. Using extensive simulations, we show that our proposed Whittle index-based beam scheduling policy significantly outperforms several scheduling policies proposed in prior work in terms of the average cost, average delay, as well as energy efficiency. I Introduction The emergence of technologies like artificial intelligence (AI), virtual reality (VR), augmented reality (AR), Internet of Things (IoT), and big data analytics, has led to the deployment of a plethora of new bandwidth-intensive applications, significantly increasing the data traffic in wireless networks [ 1 ] . Future networks are expected to handle unprecedented traffic volumes, and this exponential increase in data traffic demand has made it challenging for service providers to address the resulting bandwidth shortage [ 1 ] , [ 2 ] . While sub-6 GHz bands have limited available bandwidth, the millimeter wave (mmWave) spectrum provides abundant bandwidth and is capable of supporting data rates of several gigabits per second per user. MmWave frequencies exhibit distinct propagation characteristics compared to sub-6 GHz frequencies, e.g., mmWave communications are more susceptible to blockages, rapid channel variations, high propagation losses, and atmospheric absorption [ 3 ] . A notable advantageous aspect of the shorter wavelengths of mmWaves is that they allow for a denser packing of antenna elements within the same physical dimensions compared to sub-6 GHz frequencies. These large antenna arrays can leverage beamforming to compensate for path loss and establish communication links with adequate signal-to-noise ratios (SNR) [ 4 ] . The beams generated by arrays of antenna elements in mmWave systems are narrow, leading to focused signal transmission and creating the potential to spatially multiplex several beams, enabling simultaneous service to multiple users from a single mmWave base station (mBS). In a dense mmWave cell, the number of users associated with a mBS often exceeds the number of simultaneous beams the mBS can form, as the number of beams is constrained by the number of available radio frequency (RF) chains at the mBS [ 5 ] . Since the mBS cannot serve all its users simultaneously with the limited number of available beams, beam scheduling , i.e., selecting a subset of users to which beams will be assigned in a given time slot, is essential. At the mBS, packets arrive from the core network for transmission over the wireless channel and are queued in buffers or packet queues– typically, there are separate queues for the packets of different users. Given the packet queues at the mBS for different users, the beam scheduler at the mBS determines as to which users should be allocated beams in each time slot. Several studies have addressed the problem of beam scheduling in mmWave networks. For example, beam scheduling in multi-cell scenarios was studied in [ 5 , 6 , 7 ] , and beam-aware scheduling with the objective of saving the user equipment (UE) power in mmWave networks was studied in [ 8 ] and [ 9 ] . The beam selection problem was studied in [ 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 ] . A many-to-many beam alignment protocol was presented in [ 20 ] . The work [ 21 ] investigated joint beam allocation and scheduling, whereas [ 22 ] investigated the user selection and beam allocation problem. The problem of joint user and beam selection was analyzed in [ 23 ] ; also, [ 24 ] , [ 25 ] studied joint user association and beam scheduling in mmWave networks. A method to jointly optimize user scheduling and beam scheduling was proposed in [ 26 ] . In [ 27 ] , a joint beam allocation and relay selection scheme was proposed to mitigate blockages and enhance the total transmission rate in mmWave vehicular networks. A machine learning (ML) based beam allocation algorithm for mmWave networks was proposed in [ 28 ] . In [ 29 ] , the Dueling Double Deep Recurrent Q-Network (D3RQN) algorithm was proposed for beam allocation. Also, various packet scheduling techniques, including Longest-Queue-First (LQF) [ 30 ] , Max-Weight Scheduling (MWS) [ 31 , 32 , 33 ] , Weighted Fair Queuing (WFQ) [ 34 , 35 , 36 , 37 ] , and Random scheduling [ 38 , 32 ] , can potentially be used to address the beam scheduling problem. However, to the best of our knowledge, the powerful Whittle index technique [ 39 ] has not been used for beam scheduling in a wireless network in prior work. This is the space in which we contribute in this paper. The Whittle index, originally proposed in [ 39 ] , is a popular technique for decision-making in stochastic systems, and has been applied to effectively solve problems across a wide variety of domains, including minimizing the age of information (AoI) [ 40 , 41 , 42 , 43 ] , scheduling in wireless networks [ 44 , 45 , 46 ] , allocating jobs to processors in an egalitarian processor sharing setup [ 47 ] , user association in mmWave networks [ 48 , 49 , 50 ] , pilot allocation in massive multiple-input multiple-output (MIMO) systems [ 51 ] , beam resource allocation in radar systems for multitarget tracking [ 52 ] , etc. This paper considers downlink (DL) transmission in a single-cell mmWave network, wherein a mBS serves its associated users. Time is divided into equal slots, and at the end of every time slot, a packet arrives into the queue of a user at the mBS with a certain probability. A holding cost is incurred for the packets stored in a user’s queue at the mBS in every time slot. The number of simultaneous