PaperlayerView exact revision on arXiv

Source-linked AI summary

Metasurface-enhanced Light Detection and Ranging Technology

Renato Juliano Martins, Emil Marinov, M. Aziz Ben Youssef, Christina Kyrou, Mathilde Joubert, Constance Colmagro, Valentin Gâté, Colette Turbil, Pierre-Marie Coulon, Daniel Turover, Samira Khadir, Massimo Giudici, Charalambos Klitis, Marc Sorel, Patrice Genevet

arXiv:2204.04208v1cs.ROphysics.ins-detphysics.optics

TL;DR

LiDAR scanners face a trade-off between fast beam steering and wide field of view, limiting the simultaneous multi-zone imaging sought for robotic vision. The paper cascades a fast AOD with a passive metasurface to expand coverage, demonstrating wide-angle scanning and multi-zone ranging.

  • Problem

    LiDAR scanners are constrained by a trade-off between beam-steering speed and field of view, while robotic vision calls for simultaneous imaging across multiple zones.

    Source: Introduction · pp. 2–3

  • Method

    The system cascades a commercial AOD with a radially symmetric effective-refractive-index metasurface to steer a beam over a wide field of view.

    Source: Introduction · p. 3, Ultra-fast and high FoV metasurface… · pp. 4–5

  • Results

    150×150° field of view was demonstrated with MHz-speed beam steering, alongside high-frame-rate multi-zone ranging images.

    Source: Introduction · p. 3, Conclusion · pp. 10–11

  • Takeaways & Limitations

    The authors suggest the approach could support medium- and long-range ADAS perception, combining a central forward view with peripheral sensing for pedestrian safety.

    Source: Conclusion · p. 11

    The 3D demonstration used an indoor scene and high-reflectivity suits because the laser pulse peak power was about 10 mW.

    Source: 2D and 3D LiDAR image acquisition · p. 6

Abstract

from arXiv · show

Deploying advanced imaging solutions to robotic and autonomous systems by mimicking human vision requires simultaneous acquisition of multiple fields of views, named the peripheral and fovea regions. Low-resolution peripheral field provides coarse scene exploration to direct the eye to focus to a highly resolved fovea region for sharp imaging. Among 3D computer vision techniques, Light Detection and Ranging (LiDAR) is currently considered at the industrial level for robotic vision. LiDAR is an imaging technique that monitors pulses of light at optical frequencies to sense the space and to recover three-dimensional ranging information. Notwithstanding the efforts on LiDAR integration and optimization, commercially available devices have slow frame rate and low image resolution, notably limited by the performance of mechanical or slow solid-state deflection systems. Metasurfaces (MS) are versatile optical components that can distribute the optical power in desired regions of space. Here, we report on an advanced LiDAR technology that uses ultrafast low FoV deflectors cascaded with large area metasurfaces to achieve large FoV and simultaneous peripheral and central imaging zones. This technology achieves MHz frame rate for 2D imaging, and up to KHz for 3D imaging, with extremely large FoV (up to 150{\deg}deg. on both vertical and horizontal scanning axes). The use of this disruptive LiDAR technology with advanced learning algorithms offers perspectives to improve further the perception capabilities and decision-making process of autonomous vehicles and robotic systems.

Introduction

The paper combines a fast, narrow-field acousto-optic deflector (AOD) with a passive metasurface (MS) to expand LiDAR coverage and acquire multiple-resolution zones. The module uses the beam’s position on a radially symmetric MS to control wide-angle scanning, with 2D and 3D ranging demonstrations.

Introduction

  • Conventional LiDAR scanners face a speed–field-of-view trade-off: mechanical devices scan slowly, while fast AODs typically offer at most about 2° of field of view.The introduction also notes limited fields of view for MEMS and liquid-crystal scanners and manufacturing challenges for optical phased arrays.Source: Introduction · p. 2
  • Passive metasurfaces provide fixed beam redirection, so the paper pairs one with an active AOD rather than relying on a passive metasurface alone for real-time scanning.The introduction contrasts this approach with tunable metasurfaces that can require complex electronics and may incur metallic-building-block losses.Source: Introduction · p. 3

Ultra-fast and high FoV metasurface scanning module

  • The architecture sweeps a focused beam across a metasurface using a low-field AOD, whose position-dependent deflection expands coverage to as much as 150°.The figure depicts the synchronized laser, AOD, scanning lens, metasurface, and detector in the ranging pipeline.Source: Ultra-fast and high FoV metasurface… · p. 4, Figure 1 · p. 4
  • A circular, radially symmetric phase profile maps the beam’s radial and angular impact position to its deflected angles, enabling scanning in both azimuth and elevation.The demonstrated MS uses a radially varying effective refractive index to impart changing momentum.Source: Ultra-fast and high FoV metasurface… · p. 5
  • A 1D scan produced a 2D ranging image that located three reflectors at depths of 1.5, 2.4, and 3.5 m, validating short-range imaging to about 5 m.The reported coordinates were [-0.4 m, 1.5 m], [-0.1 m, 2.4 m], and [0.6 m, 3.5 m].Source: 2D and 3D LiDAR image acquisition · p. 5, Figure 2 · p. 6
  • The system demonstrated two-axis 150° field-of-view scanning and 3D ranging of indoor actors positioned 1.2–4.9 m away.The setup also projected Lissajous scanning profiles across the large field of view.Source: 2D and 3D LiDAR image acquisition · p. 6, Figure 3 · p. 7

Mimicking human peripheral and fovea vision with multi-zone LiDAR imaging

The system mimics central and peripheral vision by scanning two zones simultaneously: a narrow, high-resolution central view and a broad, lower-resolution peripheral view. A dual-detector scheme collects the two views, including a central image covering about 2° FoV.

Mimicking human peripheral and fovea vision with multi-zone LiDAR imaging

  • A cm-size metasurface deflector, placed after the AOD without a scanning lens, supports collimated-beam operation for multi-zone imaging.The paper describes larger metasurfaces and collimated beams as a way to reduce divergence and improve resolution.Source: Mimicking human peripheral and fovea… · p. 7
  • The 0th-order beam provides high-resolution central imaging, while the 1st-order beam scans a larger FoV at lower resolution.The two images are acquired simultaneously using a dual-detector scheme.Source: Mimicking human peripheral and fovea… · p. 8
  • Figure 4 illustrates the dual-zone setup and shows peripheral actors alongside a highly resolved, forward chessboard-like object.The caption describes the large-FoV ranging image and the narrow central view; the text identifies the peripheral actors and central target.Source: Mimicking human peripheral and fovea… · p. 8, Figure 4 · p. 8
  • The central ranging image uses the 0th-diffraction beam over a FoV of about 2°.The central detector selects the narrow-FoV return using a spatial filter; the other detector blocks the central region to collect the broader-FoV return.Source: Mimicking human peripheral and fovea… · pp. 8–9

High-speed velocimetry and time-series imaging

The system sustains MHz-rate beam deflection over extreme FoV, with a resolution trade-off at higher scanning frequencies. Time-series imaging tracks a rotating chopper and recovers the size of reflective tape from ranging data.

High-speed velocimetry and time-series imaging

  • Less than −3 dB loss persists up to around 6 MHz for single-axis scanning and 10 MHz for double-axis scanning.Correct imaging was demonstrated up to 6.25 MHz over an FoV greater than 140°; higher rates can increase frame rate while reducing resolution.Source: High-speed velocimetry and time-series… · p. 9
  • The chopper experiments used 741, 1020, and 3401 fps; all yielded an average rotation speed of 92.71 Hz, and the recovered tape size was estimated at 4 cm.The authors attribute the difference from the nominal 100 Hz to phase-jitter control and the 1 cm tape-size discrepancy to scattering from nearby screws.Source: High-speed velocimetry and time-series… · pp. 9–10, Figure 5 · p. 9
  • At a 980 µs time step, the authors estimate that a supersonic object 15 m away could cross a 120° FoV in about 74 ms and be detected in 76 time-series events.They estimate a maximum event-detection speed of 47 Mm/h using the Nyquist limit of four time-series to recover speed.Source: High-speed velocimetry and time-series… · p. 10

Experimental methodology

The setup uses an AOD to steer light onto a metasurface that expands the scan field, then measures time of flight from returned light to reconstruct ranging images.

Experimental methodology

  • AOD deflection within 49 mrad is focused onto the metasurface, which converts a 2° × 2° FoV into 150° × 150°.The metasurface is described as a designer-defined passive device; the detector monitors scattered light at each scanned angle.Source: Experimental methodology · p. 10
  • The angular scan is orchestrated in a single shot, and ranging pixels are formed by extracting time of flight from the detector signal.The method differentiates the signal, collects its peak, separates single pulses, and folds them at the scanning frequency to form an image.Source: Experimental methodology · p. 10
  • A 3 Gsample/s scope and detector rise time below about 330 ps yield an estimated single-echo depth resolution of about 5 cm.The estimate applies to single-echo, per-laser-shot measurement.Source: Experimental methodology · p. 10

Conclusion

The authors conclude that a fast deflector cascaded with a passive metasurface enables MHz beam steering across a 150° × 150° FoV. They also report multi-zone ranging and real-time time-event imaging, with possible relevance to robotic and ADAS perception.

Conclusion

  • MHz beam steering across a 150° × 150° FoV is reported to improve wide-angle scanning rate by five orders of magnitude over mechanical devices.The system also performs fast steering in one and two angular dimensions and retrieves time of flight for ranging.Source: Conclusion · pp. 10–11
  • The system demonstrates multi-zone ranging and time-event imaging above 1 kfps, reaching MHz frame rates for 1D scanning.The authors suggest possible ADAS use for central forward perception and peripheral pedestrian sensing.Source: Conclusion · p. 11

Supplementary Information

The supplementary characterization calibrates AOD voltage to deflection and measures how beam divergence and size vary with deflection angle. It identifies efficiency loss at extreme angles and increased divergence and beam waist toward the periphery.

Affiliations

  • The supplementary material includes affiliations for the contributing institutions and links to LiDAR and beam-steering videos.Source: Affiliations · p. 15

S1. Beam deflection characterization

  • AOD deflection calibration spans a 150° total FoV, but extreme angles are not observed experimentally because metasurface efficiency falls there.The reported measurements show deflections from −60° to 60° for 0–10 V; a third-order polynomial fit provides continuous calibration, while the maximum-angle limits coincide with beam deformation and efficiency drop.Source: S1. Beam deflection characterization · p. 15, Figure S1 · p. 16
  • Beam divergence rises from 1.4° at the center to 2.0° at −60°, while the beam waist grows about fourfold.The characterization covers deflection directions from −60° to +60°; a small residual zeroth diffraction order is also observed near 0°.Source: S1. Beam deflection characterization · p. 16

S2. Losses and power distribution

The supplementary section quantifies optical losses across the deflection system and describes a voltage-based calibration for two-axis scans. The calibration converts desired angular coordinates into AOD driver voltages.

S2. Losses and power distribution

  • The loss analysis includes both AOD and metasurface deflection efficiency, while the proof-of-concept system uses visible wavelengths that are not optimal for LiDAR.The authors note that the concept can be scaled to infrared frequencies.Source: S2. Losses and power distribution · pp. 16–17
  • 63% of incoming intensity is recovered in the AOD’s deflected beam; the remaining 37% leaks mainly into the 0th order and other higher orders.A beam blocker discards these unwanted signals; the deflection module’s overall loss is around 4 dB, with an additional 1.5 dB for 2D scanning.Source: S2. Losses and power distribution · p. 17

S3. Calibration of 2-axis scanning

  • For two-axis calibration, the radial-symmetry design maps the impact position on the metasurface to deflection angles, then converts the angles to AOD voltages.The voltage pair is converted to polar coordinates, and the outgoing angle is related to spherical azimuth and elevation; a polynomial fit supplies the calibration curves.Source: S3. Calibration of 2-axis scanning · pp. 17–18
  • The calibration geometry connects the metasurface-plane angle θMS and outgoing angle α to the two-axis scan coordinates.The calibration maps associate desired azimuth and elevation angles with X- and Y-driver voltages, enabling arbitrary scans.Source: Figure S1 · p. 18, Figure S2 · p. 19

S4 Beam modulation speed and imaging quality

The section tests how AOD transit time constrains fast scanning and evaluates the resulting image quality. Imaging remains usable below a measured frequency threshold, while faster scans degrade resolution.

S4 Beam modulation speed and imaging quality

  • Image quality shows no degradation below 3.33 MHz, but resolution is drastically reduced beyond that frequency.The USAF target was imaged at 1 m; the figure caption likewise reports degradation when scanning exceeds the 3 dB threshold.Source: S4 Beam modulation speed and imaging… · p. 19, Figure S3 · p. 20
  • The repointing threshold is about 6 MHz for the 2D AOD and 10 MHz for the 1D AOD, beyond the estimated transit-time limit.The threshold is defined where the repointing signal drops to around 50%.Source: S4 Beam modulation speed and imaging… · p. 19
  • The AOD’s nominal transit time is 15.4 ns, corresponding to a nominal scanning frequency of 216 kHz.The estimate uses a beam diameter of about 3 mm and acoustic velocity of 650 m/s; fast scanning beyond transit time introduces beam deformation.Source: S4 Beam modulation speed and imaging… · p. 19

S5 Beam divergence characterization

The divergence characterization measures beam profiles across ranging distances and metasurface diameters. Larger metasurfaces reduce divergence, with the stated 0.25° target requiring a 3.9 mm sample under the specified optical setup.

S5 Beam divergence characterization

  • Divergence is estimated by fitting Gaussian profiles of the beam at multiple distances and linearly regressing beam waist against distance.For the 1 mm metasurface, the reported divergence is 4.5°.Source: S5 Beam divergence characterization · p. 20, Figure S6 · p. 21
  • A 3.9 mm metasurface is indicated to reach 0.25° divergence, a value associated with resolving small objects at about 200 m.The estimate applies to a scanning-lens setup with a 3 mm beam input to the AOD; changing the optics or beam requires reconsidering divergence and calibration.Source: S5 Beam divergence characterization · p. 21, Figure S7 · p. 22
  • Increasing metasurface diameter from 1 mm through 2 mm to 3 mm reduces deflected-beam divergence while the phase function is held constant.The authors attribute the reduction to decreased spatial variation of the phase profile for a fixed beam spot.Source: S5 Beam divergence characterization · p. 21

S6. Data generation, acquisition, and analysis

This section describes the synchronized acquisition and processing used to generate LiDAR ranging images. It also demonstrates imaging of a fast-moving target, with a measured tape displacement speed of about 1400 Km/h.

S6. Data generation, acquisition, and analysis

S8. Details on the three dynamic beam scanning experiments

Three dynamic-scanning configurations compare frame rate, pixel count, and image quality while keeping the ranging limit constant. The results show a trade-off between image resolution and scanning speed.

S8. Details on the three dynamic beam scanning experiments

  • At 16.66 MHz and 150×150 pixels, imaging ran at 741 fps, but quality degraded beyond the 3 dB limit although chopper dynamics remained recoverable.The estimated tape size was slightly larger than expected.Source: S8. Details on the three dynamic beam… · p. 25
  • At 5 MHz and 70×70 pixels, the tape size was recovered correctly with good temporal resolution; 16.66 MHz and 70×70 pixels achieved the best temporal resolution, 3 kpfs.The authors describe the optimal setting as a compromise between resolution and scanning speed.Source: S8. Details on the three dynamic beam… · p. 25
  • Across the three experiments, the ranging limit was held constant and reflective-tape size was estimated.Table 1 lists frame rate, image resolution, frame time, scanning speed and laser repetition rate, measured chopper speed, and acquired-frame count.Source: S8. Details on the three dynamic beam… · p. 25, Table 1 · p. 26
  • Figure S41 presents dynamic measurements at 741 fps, 1.02 kfps, and 3.4 kfps.The caption assigns these rates to panels a, b, and c, respectively.Source: Figure S41 · p. 25

S9. Metasurface fabrication

The section describes separate fabrication routes for small GaN metasurfaces and large-area metasurfaces used in dual-field-of-view imaging. The large-area route uses nanoimprint lithography followed by etching to form GaN pillars.

S9. Metasurface fabrication

  • Small 1, 2, and 3 mm components were fabricated by growing GaN on sapphire, patterning HSQ with electron-beam lithography, and etching the GaN.The process used MOCVD for GaN growth and RIE for etching.Source: S9. Metasurface fabrication · p. 26
  • The sample image shows two metasurface deflectors, while SEM views show nanopillar size variation and phase-gradient periodicity.These images illustrate the deflector sample and its nanopillar structure.Source: Figure S52 · p. 27
  • Large-area metasurfaces for dual-FoV imaging were made by nanoimprint lithography, using a stamped resist pattern, nickel lift-off, and GaN etching to form pillars.The process uses an e-beam-fabricated master mold and an inverse textured stamp.Source: S9. Metasurface fabrication · pp. 27–28, Figure S63 · p. 28
Loading 2204.04208v1…