PaperlayerView exact revision on arXiv

Source-linked AI summary

3GPP-Compatible Long-Range Sensing for LEO-ISAC: A Window-Grid Processing Framework

Yi Geng, Xun Fan, Chenfeng Xu, Yu Zhao

arXiv:2609.30653v4cs.ITmath.IT

TL;DR

Millisecond-scale delays and multi-target receive windows make conventional OFDM sensing unsuitable for bistatic LEO-ISAC. The paper introduces receiver-side window-grid processing that preserves the 3GPP symbol structure. In simulations, the satellite-selection rule detected every target with range errors below 4 m.

  • Problem

    Conventional OFDM processing fails when TX–RX symbol misalignment and unequal transmit and receive durations occur in LEO-ISAC.

    Source: B. Why Terrestrial ISAC Processing… · p. 2

  • Method

    The WG framework processes swath-bounded echoes at the receiver without modifying the 3GPP symbol structure.

    Source: §I · p. 1, §V · p. 6

  • Results

    With the satellite-selection rule |εsa| ≤ 0.5, all targets in the simulated trials were detected with range errors below 4 m.

    Source: §V · p. 6

  • Takeaways & Limitations

    WG processing can also be reused for structured timing acquisition when synchronization is unavailable, sharing FFT outputs between communication and sensing.

    Source: D. Discussion · p. 4

    Ephemeris-aided compensation of satellite-induced Doppler shift and delay variation, which would allow longer CPIs, is left for future work.

    Source: §V · p. 6

Abstract

from arXiv · show

This paper addresses long-range sensing in bistatic low-Earth-orbit integrated sensing and communication (LEO-ISAC). Conventional OFDM sensing fails in LEO-ISAC because millisecond-scale propagation delays destroy the transmitter--receiver symbol alignment and, with multiple targets, make the receive window longer than the transmitted block. We propose a window-grid (WG) processing framework that exploits the known beam-intersection swath to resolve both problems purely at the receiver, leaving the 3GPP symbol structure unmodified: the swath delay span is folded into cyclic-prefix-sized slices, each handled by one WG. With 103.2 MHz bandwidth at bistatic ranges beyond 640 km, simulations demonstrate meter-level ranging of two targets, and 1000 Monte-Carlo trials over random target--satellite geometries show that, with an ephemeris-based satellite-selection rule, every target is detected with a 90th-percentile range error of 2.4 m.

I. INTRODUCTION

Bistatic LEO-ISAC offers wide-area sensing, but its long propagation delays make conventional OFDM sensing impractical. The paper proposes receiver-side window-grid processing that uses the known beam-intersection swath without modifying the transmitted 3GPP symbol structure.

I. INTRODUCTION

  • The bistatic setup uses a gateway transmitter and a satellite receiver, whose narrow beams overlap in an approximately 20-km-long, several-hundred-meter-wide swath.The satellite receiver is at 600 km altitude.Source: §I · p. 1
  • Propagation delays beyond 2 ms would require a cyclic prefix longer than 2 ms, far exceeding the symbol body and making conventional OFDM sensing impractical.Symbol bodies last a few to tens of microseconds.Source: §I · p. 1
  • The proposed WG framework extends sensing beyond 600 km without modifying the transmitted signal, unlike waveform-altering designs and geometry-agnostic sliding-window receivers.The paper identifies prior approaches as reaching at most several kilometers.Source: §I · p. 1

A. OFDM Signal and Echo Model

  • The sensing block uses M subcarriers across bandwidth B = M∆f and Ns OFDM symbols, each with a cyclic prefix and symbol body.The block duration is defined as CPI = NsTsym.Source: A. OFDM Signal and Echo Model · pp. 1–2
  • The target delay span is bounded by the beam-intersection swath, so a receive window of length CPI plus the swath delay span captures all target echoes.The paper defines the reference instant as the earliest possible echo arrival.Source: A. OFDM Signal and Echo Model · p. 2

B. Why Terrestrial ISAC Processing Fails for LEO-ISAC

Terrestrial OFDM sensing depends on transmitter–receiver symbol alignment and equal transmit and receive block durations. Long LEO-ISAC delays and multi-target swath coverage violate these conditions, preventing direct use of that processing.

B. Why Terrestrial ISAC Processing Fails for LEO-ISAC

  • Misaligned received and transmitted symbols make element-wise division produce pseudorandom data, scattering echo energy across the range-Doppler map instead of forming a discernible peak.The scattered energy merely raises the noise floor.Source: B. Why Terrestrial ISAC Processing… · p. 2
  • Capturing all targets requires a receive window of length CPI plus the swath delay span, longer than the transmitted block and incompatible with element-wise division.Source: B. Why Terrestrial ISAC Processing… · p. 2
  • The paper states that conventional terrestrial OFDM-ISAC processing cannot be applied directly and that, to its knowledge, no OFDM sensing scheme had addressed both LEO-ISAC problems.Source: B. Why Terrestrial ISAC Processing… · p. 2

III. PROPOSED WINDOW-GRID PROCESSING

The framework uses known beam-intersection-swath arrival bounds to process echoes entirely at the receiver while preserving the 3GPP symbol structure.

III. PROPOSED WINDOW-GRID PROCESSING

  • The WG framework uses ephemeris- and beam-geometry-known arrival bounds to operate at the receiver without modifying the 3GPP symbol structure.Source: §III · p. 2

A. Transmit–Receive Timing

The receiver anchors its timing at the earliest possible echo and folds the absolute propagation delay into a relative delay set by the swath. Window grids then replicate the transmitted symbol-body timing across the receive sequence.

A. Transmit–Receive Timing

  • Anchoring reception at the earliest possible echo and adding a silent guard interval equal to the swath delay span captures all target echoes without overlap from the next block.The receive window records NsNsym + ⌈∆τswfs⌉ samples.Source: A. Transmit–Receive Timing · p. 2
  • The timing reference folds millisecond-scale absolute delay into a microsecond-scale relative delay, so processing handles the swath span rather than the full propagation delay.Source: A. Transmit–Receive Timing · p. 2

1) WG pattern:

  • Each WG has Ns body-length windows replicating the transmitted symbol-body timing, with CP-length gaps; its anchor is the start of the CP preceding the first window.The figure illustrates WG pattern, capture, and processing.Source: 1) WG pattern · pp. 2–3, Figure 2 · p. 3

2) WG placement:

The WG grid shifts a symbol-timed window pattern across the receive sequence in cyclic-prefix-sized steps. Each WG covers one bistatic-range slice, and the swath delay span is partitioned across designated WGs.

2) WG placement:

  • Successive WGs shift their anchors by Ncp samples, with each anchor at tref + iTcp.The windows within each WG replicate the transmitted symbol-body timing.Source: 2) WG placement · p. 3
  • Each WG's range slice corresponds geometrically to a spatial cell bounded by the swath and two isorange ellipsoids.The cells cover bistatic ranges [Rmin + iRcp, Rmin + (i+1)Rcp).Source: 2) WG placement · p. 3
  • The swath delay span is partitioned into cyclic-prefix-length slices, each handled by one WG.Each WG senses a bistatic-range interval of width Rcp = cTcp; the unique WG matching a target's relative delay is its designated WG.Source: 2) WG placement · p. 3

3) WG processing:

Each WG processes symbol-body segments from its windows to form a range-Doppler map, with the designated WG capturing the target echo coherently. Other WGs can produce weaker ghost peaks, while satellite-induced Doppler can reduce coherent gain and cause ICI.

3) WG processing:

  • The Ns segments captured by a WG's windows are OFDM-demodulated and stacked column by column to form its processing input.Element-wise division by the transmitted subcarriers produces the channel matrix before the 2D FFT generates the WG's range-Doppler map.Source: 3) WG processing · p. 3
  • Satellite-induced Doppler reduces coherent gain by |Γ(ε)| and leaks a fraction 1 − sinc^2(ε) of echo power into ICI.Both effects are negligible for |ε| ≪ 1; εsa can be predicted from ephemeris and the WG cell geometry to support satellite selection.Source: 3) WG processing · p. 3
  • The designated WG captures the full Ndata samples and yields the true peak with full coherent gain, while undesignated WGs may produce weaker ghost peaks.Early and late windows capture only part of the echo symbol; a 942-sample relative delay designates WG 15 in the illustrated example.Source: 3) WG processing · pp. 3–4
  • The inverse FFT makes range estimates periodic with period Runa, the maximum unambiguous bistatic range.The range-Doppler map is formed by applying a 2D FFT to the channel matrix.Source: 3) WG processing · pp. 3–4

C. Ghost-Peak Analysis and Suppression

Ghost peaks are suppressed by cropping each WG's range-Doppler map to the cyclic-prefix-sized valid range and relying on the surviving peaks' lower amplitude. At most one ghost per side survives cropping, and the remaining peaks are attenuated by about 23 dB for the stated 3GPP numerology.

C. Ghost-Peak Analysis and Suppression

  • Cropping retains only the first Ncp bins of each RD map, where the target's designated WG contains the true peak.Each WG is designed for relative ranges in [0, Rcp), whereas the map spans [0, Runa) and Runa ≫ Rcp.Source: C. Ghost-Peak Analysis and Suppression · p. 4
  • At most one ghost peak per side survives cropping, at WGs approximately floor(Ndata/Ncp) steps from the designated WG.Early-side ghosts move out of the valid region; late-side ghosts wrap, and only peaks meeting the stated modulo-range intervals survive.Source: C. Ghost-Peak Analysis and Suppression · p. 4
  • Ghost-peak amplitude decreases linearly with distance from the designated WG because the captured echo length falls by Ncp per WG.This monotonic decay underlies the attenuation of the surviving ghosts.Source: C. Ghost-Peak Analysis and Suppression · p. 4
  • 23 dB attenuation relative to the true peak is expected for surviving ghosts with 3GPP normal-CP numerology at 6.6% CP overhead.In high-path-loss LEO-ISAC, this typically pushes surviving ghosts below the noise floor; visible survivors can be discarded using their shared range rate and predicted WG locations.Source: C. Ghost-Peak Analysis and Suppression · p. 4

D. Discussion

The WG branches operate independently and can be distributed across parallel processors. Reusing WG processing when synchronization is unavailable can also provide a structured timing-acquisition search and a common receiver structure for communication and sensing.

D. Discussion

  • Each WG requires Ns OFDM demodulations and one 2D FFT, and the total complexity scales linearly with Ng.The WG branches are mutually independent and can be distributed across processing units such as multi-core CPUs for real-time requirements.Source: D. Discussion · p. 4
  • When synchronization is unavailable, WG processing can be reused as a structured timing-acquisition search.Its unmodified 3GPP symbol structure and shareable FFT outputs provide a common receiver structure for communication and sensing.Source: D. Discussion · p. 4

A. Simulation Setup

The simulations use a 107-WG swath and two aircraft to test ranging, multi-target processing, and ghost-peak suppression. The designated WGs detect both targets at meter-level errors, while wrapped ghost peaks are attenuated below the noise floor.

A. Simulation Setup

  • The 107 WGs span a swath from 644.236 to 663.186 km; the 3GPP-compatible 120 kHz numerology gives each WG a 177.2 m range span.Two aircraft at 663.0662 km and 646.9711 km fall in WGs 106 and 15, respectively.Source: A. Simulation Setup · p. 5
  • With satellite velocity (7.5, 0, 0) km/s, εsa was 0.5 and 0.36 for the two targets, versus −0.83 and 14 dB loss with velocity (0, 7.5, 0) km/s.The authors select a sensing satellite with |εsa| ≤ εth using swath geometry and ephemeris before detection.Source: A. Simulation Setup · p. 5

B. Simulation Results

  • 2.4 m and 1.6 m range errors were measured for targets in WGs 106 and 15, respectively, and both targets were detected despite millisecond-scale delays and echoes extending beyond the CPI.The errors are within one range resolution, c/B.Source: B. Simulation Results · p. 5
  • Measured range rates include both satellite- and target-induced contributions; recovering the target-induced component requires ephemeris-aided compensation and remains future work.Source: B. Simulation Results · p. 5
  • Ghost peaks in WGs 10 and 5 appeared at 959.7 m and 1841.9 m, close to predicted positions, then shifted away from the valid region and were removed by cropping.Source: B. Simulation Results · p. 5
  • In WG 1, a wrapped ghost peak captured only 40 correct samples and was attenuated by 26.6 dB, leaving it below the noise floor.Its 21-sample relative delay lay inside the valid region despite the attenuation.Source: B. Simulation Results · p. 5

C. System-Level Performance

Across 1000 random target–satellite geometries, satellite selection using |εsa| ≤ 0.5 yielded detections in every admitted trial and a 2.4 m 90th-percentile range error. The conclusion reports errors below 4 m for all selected trials and identifies ephemeris-aided compensation as future work.

C. System-Level Performance

  • The 1000 trials used random target–satellite geometries, ideal synchronization and ephemeris, and no modeled clutter; misses counted as infinite range errors.The per-map false-alarm probability was set to 10−2.Source: C. System-Level Performance · pp. 5–6
  • Figure 5 shows range-Doppler maps plotting relative range against bistatic range rate for WGs 106, 15, 10, 5, and 1, in panels (a)–(e), respectively.Source: Figure 5 · p. 6
  • A satellite-selection rule |εsa| ≤ 0.5 yielded detections in all 263 admitted trials, with a 2.4 m 90th-percentile and 3.8 m maximum range error.The rule admitted 26% of drawn satellites; Fig. 6(a) shows the range-error CDF and Fig. 6(b) relates Pd and coherent loss to |εsa|.Source: C. System-Level Performance · p. 6, Figure 6 · p. 6

V. CONCLUSION

  • The framework resolves TX–RX symbol misalignment and unequal TX and RX durations at the receiver while keeping the symbol structure 3GPP-compatible.The swath delay span is folded into CP-sized slices, each handled by one WG.Source: §V · p. 6
  • The conclusion reports that all targets in simulated trials using |εsa| ≤ 0.5 were detected with range errors below 4 m.Ephemeris-aided compensation of satellite-induced Doppler shift and delay variation, which would allow longer CPIs, remains future work.Source: §V · p. 6
Loading 2609.30653v4…