PaperlayerView exact revision on arXiv

Source-linked AI summary

Low Overhead IMU Assisted Predictive Beam Management for Multiband LEO Direct to Device Links

Abdulrahman Al Hababi, Meysam Ghanbari, Mohammad Taghi Dabiri, Mazen O. Hasna, Khalid A. Qaraqe

arXiv:2610.01635v1eess.SP

TL;DR

Ku-band handheld LEO access needs directional gain across multiple handset panels, but the large candidate set and fast hand motion complicate beam management. The paper predicts handset orientation from delayed IMU data, trains a small diverse candidate set, and activates Ku band only when its predicted rate exceeds the anchor by a margin. With six pilots at 90◦/s, it reports higher goodput and lower outage than an equal-budget delayed-attitude baseline.

  • Problem

    A four-panel handset with two visible satellites has 392 Ku-band candidates, while ordinary motion can change the best beam within a decision interval.

    Source: §I · p. 1

  • Method

    The scheme predicts current orientation from delayed IMU data, selects a small satellite/panel-diverse pilot set, and activates Ku band only when its predicted post-training rate exceeds the anchor by a margin.

    Source: §I · p. 1, B. Contributions · p. 2

  • Results

    With six pilots (0.3% overhead) at 90◦/s, goodput improves by 21.2% and broadband outage falls by 57.6% versus an equal-budget delayed-attitude baseline.

    Source: B. Contributions · p. 2

  • Takeaways & Limitations

    With six pilots, the method preserves approximately 81% instantaneous-best-beam capture across the tested angular speeds.

    Source: §V · p. 6

    The panel patterns and blockage process are analytical; replacing them with measured handset data remains future work.

    Source: §V · p. 6

Abstract

from arXiv · show

Direct to device D2D connectivity from low Earth orbit LEO satellites is moving to Ku band, where a handheld terminal must obtain directional gain from several small phased array panels distributed around the chassis. Beam management then becomes a joint satellite panel beam selection problem with hundreds of candidates, and ordinary hand motion can change the best candidate within one decision interval. This paper proposes a low overhead predictive beam management scheme for a multiband LEO D2D downlink in which a low frequency anchor link carries control signalling and fallback traffic, and a Ku band link carries broadband data. The handset inertial measurement unit IMU reports attitude and angular rate with a known delay; the scheme extrapolates the delayed attitude to the current orientation, scores all satellite panel beam candidates analytically from the satellite ephemeris and panel geometry, and trains only a small candidate set built with a satellite panel diversity rule under a fixed pilot budget. The Ku band link is activated only when its predicted post training rate exceeds the anchor rate by a margin. Trace driven Monte Carlo simulations with a four panel handset and two visible satellites show that, with six pilots 0.3 percent training overhead, the proposed scheme improves mean goodput by 21.2 percent and reduces broadband outage by 57.6 percent at 90 degrees per second relative to an equal budget delayed attitude baseline, and operates within 3.8 percent of a zero overhead oracle.

I. INTRODUCTION

Ku-band handheld LEO links require directional gain from multiple handset panels, making beam selection costly and sensitive to hand motion. The paper combines predictive, diversity-aware candidate selection with a low-band fallback and reports gains over an equal-budget delayed-attitude baseline.

I. INTRODUCTION

  • For a four-panel terminal and two visible satellites, exhaustive search spans 392 candidates and nearly 20% of a 100 ms interval, while ordinary motion can change the best beam rapidly.The link-budget need for directional gain motivates distributed panels; residual uncertainty is primarily handset attitude because the satellite line-of-sight direction is known from ephemeris.Source: §I · p. 1

A. Related Work

  • Prior sensor-aided methods target terrestrial links or fixed and vehicular terminals; the paper identifies joint multi-satellite, multi-panel selection, delayed-IMU prediction, and low-band fallback as unaddressed together.The related-work discussion also notes that some learning-based methods require site-specific training data.Source: A. Related Work · p. 2

B. Contributions

  • A low-frequency anchor carries control signalling and fallback traffic, while the multi-panel handset uses Ku band for broadband data.The model includes quaternion-based attitude, panel responses, panel-dependent Markov hand/head blockage, and pilot overhead in effective goodput.Source: B. Contributions · p. 2
  • The diversity-constrained scheme selects an L-pilot set so pilots are not concentrated on adjacent codewords from a single panel.It analytically scores satellite–panel–beam candidates after extrapolating delayed IMU attitude; the supplied passage truncates the description of this step.Source: B. Contributions · p. 2
  • With six pilots (0.3% overhead) at 90◦/s, the scheme improves goodput by 21.2% and reduces broadband outage by 57.6% versus an equal-budget delayed-attitude baseline.The comparison uses trace-driven Monte Carlo simulation.Source: B. Contributions · p. 2

II. SYSTEM MODEL

The system model pairs a robust low-frequency anchor with a Ku-band broadband link and represents each Ku-band option as a satellite, handset panel, and beam. It uses handset attitude to express satellite direction and beam mismatch in handset coordinates.

A. Multiband LEO Direct-to-Device Architecture

  • The low-frequency link provides control signalling and fallback service, while Ku band provides broadband data; either service may be provided by any visible satellite.The illustrated serving links are examples, not permanent satellite assignments.Source: A. Multiband LEO Direct-to-Device… · p. 2
  • A Ku-band candidate is indexed by satellite, panel, and beam; the simulated four-panel handset has 392 candidates across two visible satellites.Each panel has a 7 × 7 azimuth–elevation codebook, giving 196 panel–beam pairs per satellite.Source: A. Multiband LEO Direct-to-Device… · p. 3
  • The handset attitude is represented by a unit quaternion, which transforms the satellite line-of-sight into handset coordinates for calculating beam mismatch.The model also specifies satellite slant range from elevation angle, Earth radius, and satellite altitude.Source: A. Multiband LEO Direct-to-Device… · p. 3

B. Panel Gain, Blockage, and Link Rate

The link model combines array, panel, polarization, and blockage effects to determine candidate gain and received power. It also accounts for atmospheric loss and the pilot cost of selection when calculating effective goodput.

B. Panel Gain, Blockage, and Link Rate

  • Handset gain combines the steered-array pattern, panel element response, polarization mismatch, and panel-dependent blockage.The auxiliary loss combines projected copolarization loss and hand/head blockage; blockage is modeled per panel by a two-state Markov process.Source: B. Panel Gain, Blockage, and Link Rate · p. 3
  • The front-facing panel has an element contribution, while a fixed front-to-back attenuation applies when its normal faces away from the satellite direction.The supplied passage states the condition for applying the attenuation but does not provide its value.Source: B. Panel Gain, Blockage, and Link Rate · p. 3
  • The model includes an elevation-dependent atmospheric term in received power, consistent with elevation scaling used in Earth–space link budgets.The passage identifies zenith attenuation and rain as components of the atmospheric term.Source: B. Panel Gain, Blockage, and Link Rate · p. 3
  • Effective goodput accounts for pilot training time after candidate selection, while the low-band anchor prevents complete service interruption when predicted Ku-band service is unreliable.The candidate rate is calculated before training overhead is applied.Source: B. Panel Gain, Blockage, and Link Rate · p. 3

III. IMU-ASSISTED DIVERSITY-AWARE PREDICTIVE BEAM MANAGEMENT

The beam manager addresses rapid changes in the best satellite–panel–beam option by predicting current orientation from delayed attitude and angular-rate measurements. It uses that prediction to restrict training to a small, panel-diverse set.

III. IMU-ASSISTED DIVERSITY-AWARE PREDICTIVE BEAM MANAGEMENT

  • The best satellite–panel–beam combination can change within the IMU reporting delay, making exhaustive search costly and neighbouring-beam tracking unreliable under rapid rotation.An exhaustive sweep requires one pilot per candidate per decision.Source: §III · p. 3
  • The scheme predicts current handset orientation from delayed attitude and angular-rate measurements.The prediction targets the residual uncertainty in handset attitude.Source: §III · p. 3
  • It restricts training to a small, panel-diverse candidate set rather than sweeping every candidate or tracking only neighbours of the previous choice.This addresses the high overhead of exhaustive search and the unreliability of local tracking during rapid rotation.Source: §III · p. 3

A. Attitude Prediction and Candidate Scoring

The method predicts current handset orientation from delayed attitude and angular-rate measurements, then analytically scores candidate rates. It uses satellite–panel grouping to preserve diversity when selecting candidates for training.

A. Attitude Prediction and Candidate Scoring

  • Delayed attitude and body angular velocity feed a first-order quaternion prediction of the handset’s current orientation.The delay models the age of the estimate when the sounding set is committed.Source: A. Attitude Prediction and Candidate… · p. 3
  • Predicted candidate rates are calculated using the predicted attitude, panel-blockage loss, and a conservative SINR margin.The candidate rates are obtained from the paper’s earlier geometry and response equations.Source: A. Attitude Prediction and Candidate… · p. 3
  • Because globally strongest candidates can cluster on one panel, the method partitions candidates into satellite–panel groups and identifies each group’s best predicted codeword.This addresses blockage of a dominant panel and attitude errors that cross panel boundaries.Source: A. Attitude Prediction and Candidate… · pp. 3–4
  • The scheme ranks group winners and assigns pilots across the strongest groups, preserving satellite/panel diversity within the fixed budget.Near a panel boundary, this ensures the alternate panel is sounded, leaving the miss probability bounded by within-panel codeword error.Source: A. Attitude Prediction and Candidate… · p. 4

B. Pilot Selection, Activation, and Fallback

Only the selected candidate set is sounded, so radio-training cost depends on the pilot budget rather than the full codebook. Ku-band transmission is used only when the predicted post-training rate clears the low-band fallback rate by a margin.

B. Pilot Selection, Activation, and Fallback

  • Ku-band transmission is activated only when its predicted post-training rate exceeds the low-band rate by the activation margin; otherwise, the terminal stays on the low-band link.The rule avoids spending data time on a predicted Ku-band opportunity that cannot exceed the fallback rate after training.Source: B. Pilot Selection, Activation, and… · p. 4
  • Only candidates in the selected set are sounded, making online radio-training cost L pilots rather than the full candidate count |C|.Candidate scoring still takes O(|C|) lightweight arithmetic operations based on geometry and IMU data.Source: B. Pilot Selection, Activation, and… · p. 4

A. Simulation Setup

The evaluation uses trace-driven Monte Carlo simulation with two LEO satellites and a four-panel handset. It varies handset angular speed while holding orientation excursion fixed, and compares the six-pilot search with higher-overhead alternatives.

A. Simulation Setup

  • The trace-driven Monte Carlo simulator models two LEO satellites at 550 km, using a 14 GHz Ku-band carrier and 200 MHz bandwidth.The setup also specifies satellite EIRP of 83 dBm, a 6 dB receiver noise figure, 2 dB implementation loss, and −3 dB aggregate interference-to-noise ratio.Source: A. Simulation Setup · p. 4
  • The simulation sets the low-band fallback rate to 8 Mbps, the broadband threshold to 50 Mbps, and the activation margin to µ = 1.10.Spectral efficiency is capped at 8 bit/s/Hz, and the predictive SINR margin is 2.5 dB.Source: A. Simulation Setup · p. 4
  • The four-panel handset uses a 100 ms decision period and 50 µs pilots; six predictive pilots consume 0.3% overhead, versus 6.76% for hierarchical search and 19.6% for an exhaustive sweep.The panels have 12 dBi maximum gain, 18° half-power beamwidth, and 15° codebook spacing.Source: A. Simulation Setup · p. 4

B. Goodput and Training Overhead

With six pilots, the proposed method improves goodput over an equal-budget delayed-attitude baseline at high handset angular speed, while using substantially less overhead than alternative search methods. Its candidate capture probability is lower than hierarchical sweeping’s at 90°/s, but remains close to 81% at 45 and 90°/s.

B. Goodput and Training Overhead

  • At 90°/s, goodput reaches 136.20 Mbps, 21.2% above delayed attitude and 35.8% above reactive tracking; it is 3.8% below the zero-overhead oracle.At 45°/s, goodput is 132.35 Mbps, versus 125.64 Mbps for delayed attitude and 125.11 Mbps for hierarchical sweeping.Source: B. Goodput and Training Overhead · p. 4
  • The proposed 0.3% overhead is 86.5% lower than reactive tracking, 95.6% lower than hierarchical sweeping, and 98.5% lower than exhaustive sweeping.At 90°/s, hierarchical sweeping captures the instantaneous best beam more often than the proposed method (88.6% versus 80.7%) while sounding more than twenty times as many candidates.Source: B. Goodput and Training Overhead · p. 5
  • Fig. 4 compares goodput and overhead against the number of predictive pilot candidates at 45°/s.The supplied caption identifies the comparison dimensions but does not state the plotted outcome.Source: Figure 4 · p. 5

C. Broadband Outage

The six-pilot method improves goodput over an equal-budget delayed-attitude baseline and approaches the zero-overhead oracle. The section reports broadband outage at 90◦/s and explains the low-overhead candidate budget choice.

C. Broadband Outage

  • At 90◦/s, broadband outage is 0.0644 versus 0.1521 for delayed attitude and 0.2955 for reactive tracking, reductions of 57.6% and 78.2%, respectively.The outage is also lower than hierarchical sweeping despite using fewer than one twentieth of its pilot budget.Source: C. Broadband Outage · p. 5

D. Predictive Pilot-Budget Tradeoff

  • At 45◦/s, moving from one to two candidates produces the largest initial goodput improvement; gains then change slowly over L = 4–8 as overhead rises 0.05L%.The initial improvement is attributed to introducing satellite/panel diversity.Source: D. Predictive Pilot-Budget Tradeoff · p. 5
  • L = 6 is a deliberately low-overhead operating point rather than the absolute goodput optimum, and increasing the budget to L = 20 provides no further gain after pilot time.Future work includes measured handset patterns and blockage, correlated hand/head blockage, longer IMU latency, satellite handover, and experimental hardware validation.Source: D. Predictive Pilot-Budget Tradeoff · p. 6, §V · p. 6

V. CONCLUSION AND FUTURE DIRECTIONS

  • At 90◦/s, six pilots (0.3% overhead) improve goodput by 21.2% over equal-budget delayed attitude and operate within 3.8% of the zero-overhead oracle.The conclusion reports that the method preserves approximately 81% instantaneous-best-beam capture across the tested angular speeds.Source: §V · p. 6
Loading 2610.01635v1…