PaperlayerView exact revision on arXiv

Source-linked AI summary

CommRad: Context-Aware Sensing-Driven Millimeter-Wave Networks

Ish Kumar Jain, Suriyaa MM, Dinesh Bharadia

arXiv:2407.08817v1eess.SPcs.NI

TL;DR

Mobile mmWave links need directional beams maintained as users move, but radar lacks the context to identify which observed objects matter for beam management. CommRad combines radio sensing with radar tracking; its evaluation reports a 2.5x median throughput improvement.

  • Problem

    Mobile mmWave links need directional beams maintained as users move, but radar lacks the context to identify which observed objects matter for beam management.

    Source: §1 · pp. 1–2, §3 · p. 4

  • Method

    CommRad combines bi-static radio sensing during beam training with mono-static radar, using radio-derived context for radar tracking and beam tracking and blockage prediction.

    Source: §1 · p. 2

  • Results

    2.5x median throughput improvement and 8x worsecase 20th percentile throughput improvement, with communication overhead as low as 0.5%.

    Source: §1 · p. 3

  • Takeaways & Limitations

    CommRad improves mmWave link quality for mobile users under environmental vulnerabilities including blockages and mobility-related beam misalignment.

    Source: §7 · p. 12

    CommRad provides base-station-side directional tracking assuming a quasi-omni user-side beam, a typical setting for antenna-constrained handhelds.

    Source: §6 · p. 12

Abstract

from arXiv · show

Millimeter-wave (mmWave) technology is pivotal for next-generation wireless networks, enabling high-data-rate and low-latency applications such as autonomous vehicles and XR streaming. However, maintaining directional mmWave links in dynamic mobile environments is challenging due to mobility-induced disruptions and blockage. While effective, the current 5G NR beam training methods incur significant overhead and scalability issues in multi-user scenarios. To address this, we introduce CommRad, a sensing-driven solution incorporating a radar sensor at the base station to track mobile users and maintain directional beams even under blockages. While radar provides high-resolution object tracking, it suffers from a fundamental challenge of lack of context, i.e., it cannot discern which objects in the environment represent active users, reflectors, or blockers. To obtain this contextual awareness, CommRad unites wireless sensing capabilities of bi-static radio communication with the mono-static radar sensor, allowing radios to provide initial context to radar sensors. Subsequently, the radar aids in user tracking and sustains mobile links even in obstructed scenarios, resulting in robust and high-throughput directional connections for all mobile users at all times. We evaluate this collaborative radar-radio framework using a 28 GHz mmWave testbed integrated with a radar sensor in various indoor and outdoor scenarios, demonstrating a 2.5x improvement in median throughput compared to a non-collaborative baseline.

1 Introduction

CommRad addresses the overhead and scalability of beam training and radar’s lack of context by combining radio sensing with radar tracking. Its evaluation reports throughput gains while maintaining mobile links through blockages.

1 Introduction

  • 5G NR beam training can consume up to 25% overhead, and its per-user feedback demands limit scalability in multi-user scenarios.The cited protocol example repeats 5 ms of beam training every 20 ms.Source: §1 · p. 1
  • Radar returns lack context to distinguish active users, useful reflectors, and blockers; prior radar-aided systems were limited to clutter-free, single-user settings without reflected-path support.The introduction says prior work had not experimentally demonstrated context-aware multi-user tracking with dynamic reflected-path support and blockage adaptation.Source: §1 · p. 2
  • CommRad uses mandatory radio beam training to provide radar with context, then relies on context-aware radar tracking during communication to maintain beams.The framework combines bi-static radio sensing with mono-static radar sensing.Source: §1 · p. 2
  • Radio-derived CSI supplies user angles and distances and reflector location, orientation, and size, allowing radar to distinguish relevant objects from its surroundings.The base station transmits a known preamble across multiple beams, and algorithms use two-dimensional CSI to estimate attributes and map them to radar’s view.Source: §1 · p. 2
  • CommRad tracks direct and reflected path angles and predicts blockages, enabling beam switching to an unblocked path when the direct path is obstructed.Tracking uses spatial-temporal consistency and Kalman filtering to reduce noise.Source: §1 · pp. 2–3
  • 2.5x median throughput and 8x worst-case 20th-percentile throughput gains, with communication overhead as low as 0.5%, are reported for CommRad.The evaluation uses a 28 GHz mmWave radio, a 24 GHz radar, and four mobile users.Source: §1 · p. 3

2 Background and Related Work

Prior work uses radar, other sensors, beam-training reductions, reflector models, and blockage methods for mmWave links, but the cited approaches leave gaps in context, mobility, or system integration.

Radar-aided mmWave systems

  • Prior radar-aided mmWave systems generally use end-to-end learning from radar images, track direct paths, and assume a single user; the cited work also lacks temporal tracking and beam-scan integration.The related-work section positions CommRad as integrating collaborative learning, user tracking, reflector modeling, and blockage mitigation.Source: §2 · p. 3

Other sensors-aided mmWave systems

  • LiDAR- and camera-aided beam tracking may require environment-specific retraining and labor-intensive ground-truth collection, while other cited sensors target less common indoor WiFi settings.The authors say CommRad’s contextual-awareness approach could be applied to other sensing modalities.Source: §2 · p. 3

High-overhead beam training

  • Beam-training methods that scan fewer beams trade reduced overhead for higher misdetection risk, while site-specific beam sets do not generalize to arbitrary cellular deployments.The section highlights these limits particularly for multi-user mobility scenarios.Source: §2 · p. 3

Reflector Modeling

  • Cited reflector models use radio data at multiple static locations for one-time deployment modeling, unlike CommRad’s real-time mobile-link goal using collaborative radar and radio learning.Source: §2 · p. 3

Blockage Prediction

  • Prior blockage approaches include prediction, stationary-obstacle handling, and recovery from permanent blockages; CommRad targets dynamic blockers and integrates prediction with contextual direct/reflected-path tracking.Source: §2 · p. 3

3 CommRad Design

CommRad combines radio sensing with radar tracking because radar detects environmental reflections but cannot identify which belong to users, reflectors, or blockers. Radio-derived context lets radar support beam management between beam-training scans.

3 CommRad Design

  • CommRad uses a three-stage process to define radar context, acquire it periodically through radio beam training, and use radar to track paths and detect blockers when beam training is unavailable.Tracking reflected paths alongside direct paths can support constructive multi-beam techniques for link reliability.Source: §3 · p. 4

3.1 What is the Context that Radar Lacks?

  • Radar context comprises active users’ approximate distance and angle, reflectors’ location, orientation, and size, and blockers’ relative location, size, and speed.These parameters support user and reflector identification, beam maintenance, and estimation of blockage timing and duration.Source: §3.1 · p. 4

How Radar Lacks Context for Users?

  • When two users cross, radar without context can incorrectly continue both along their previous trajectories; periodic radio context maps trajectories to user IDs.The described example has users moving in opposite directions across one another.Source: How Radar Lacks Context for Users? · pp. 4–5

How Radar Lacks Context for Reflectors?

  • Radar RSS alone cannot identify useful communication reflectors: its strongest response occurs at 60° orientation, while the reflector best serving the user is at 0°.For other orientations, the radio RSS diminishes; strong radar reflection therefore does not imply a useful reflected communication path.Source: How Radar Lacks Context for Reflectors? · p. 5

3.2 Acquiring Context through Radar+Radio Collaboration

CommRad uses radio measurements to supply radar with user and reflector context. For reflectors, it estimates path and surface parameters from beam scans, then refines radio-derived distance using radar.

Acquiring Context for Users

  • Radio beam scans estimate each active user’s angle from the best-RSS beam and distance from RSS, then map the estimated location into radar’s distance-angle view.The described initial scenario excludes reflectors and blockages.Source: Acquiring Context for Users · p. 5

Acquiring Context for Reflectors

  • CommRad focuses on surface reflectors, representing each as a line segment with location, orientation, and size, which are needed to compute reflected-path angles.Point reflectors are described as less suited to sustaining mobile links than surface reflectors.Source: Acquiring Context for Reflectors · p. 6
  • Subspace MUSIC separates direct and reflected paths across angle, and MUSIC across bandwidth estimates relative ToF; the shorter-ToF path is treated as direct.Path angles and distances are then determined using relative ToF and the user location obtained through RSS measurements.Source: Acquiring Context for Reflectors · p. 6
  • Reflector orientation comes from the tangent to a user-and-base-station-foci ellipse; size is learned from scans at multiple user locations and can be repeated if configuration changes.The reflector position locus lies on the ellipse, and the learned endpoints define its line segment.Source: Acquiring Context for Reflectors · p. 6
  • Because RSS-based distance can be inaccurate in multipath, CommRad uses radio for coarse estimates and refines them by finding the nearest object in radar’s view.The passage states that radar distance measurements from synchronized Tx and Rx time-of-flight measurements are more accurate.Source: Acquiring Context for Reflectors · p. 6

3.3 Context-Aware Beam Management using Radar Sensor

Once radio-radio collaboration gives radar context, CommRad removes static clutter and tracks users through radar frames while the radio is busy communicating. The tracking supports direct and reflected paths.

3.3 Context-Aware Beam Management using Radar Sensor

  • With 200 MHz bandwidth, CommRad has 0.75 m range resolution; angular resolution is 7.16° for a 16-virtual-antenna system and 14.32° for an 8-antenna system.The passage links the virtual antenna array’s improved angular resolution to better localization and tracking accuracy.Source: §3.3 · p. 7

Static Clutter removal with frame subtraction

  • Frame subtraction removes static environmental clutter from Range-AoA profiles before radio data localizes users in the clutter-removed image.The method averages multiple chirps according to mobility conditions, then subtracts the current frame’s profile from the averaged profile.Source: Static Clutter removal with frame… · p. 7, Figure 9 · p. 7

Direct path beam tracking

  • Radar tracks direct and reflected paths while radio resources are occupied; direct-path tracking follows the strongest signal within an iteratively adjusted user bounding box.Periodic radio location updates can help maintain radar tracking accuracy, especially when a user is too close for radar to retain context.Source: §3.3 · p. 6, Direct path beam tracking · p. 7, Figure 11 · p. 7

Reflected path beam tracking

CommRad tracks reflected-path beam angles by modeling the reflector as a mirror and the base station as a virtual point behind it. For blockage mitigation, radar detects blockers in the link region and the base station can switch to an alternate reflected beam.

Reflected path beam tracking

  • Reflector location, size, and orientation determine a virtual base-station position behind the mirror, allowing the reflected-path angle to be tracked as the user moves.The line from this virtual position to the user defines the reflected-path angle.Source: Reflected path beam tracking · pp. 7–8
  • The estimated change in reflected-path angle provides an alternate beam when the direct path is blocked or unavailable.Figure 12 depicts blockage tracking and proactive prevention.Source: Reflected path beam tracking · p. 8, Figure 12 · p. 8

Context-Aware Blockage Mitigation using Radar

  • A blockage-prone region is a rectangle between the user and base station; an object entering it can cause blockage.Radar tracks object velocity over frames to estimate when the blocker arrives.Source: Context-Aware Blockage Mitigation… · p. 8

24 GHz Radar

The system estimates blocker arrival and blockage duration from radar-tracked motion, then uses the estimated event duration to guide beam switching.

24 GHz Radar

  • The platform includes synchronized radio and radar units at the base station.Figure 13 shows the base-station Radio and Radar setup.Source: Figure 13 · p. 8
  • Blocker length and velocity determine the estimated duration of a blockage.The supplied passage expresses the duration as blocker length divided by blocker velocity.Source: §24 · p. 8
  • For a longer blockage, the base station switches to a reflected beam until the event ends, then returns to the original link.The base station is alerted to the upcoming event using the blocker velocity and duration.Source: §24 · p. 8

4 Implementation

The implementation integrates software-defined radios, phased arrays, and radar into a synchronized platform for indoor and outdoor experiments. The radio operates at 28 GHz, while the radar uses a 24 GHz FMCW setup.

4 Implementation

  • The platform is designed for indoor and outdoor experiments, with radio and radar subsystems integrated into a joint synchronized setup.The scenarios include four mobile users and varied reflectors such as concrete walls, monitor screens, and glass walls.Source: §4 · p. 8, Figure 14 · p. 9

4.1 Radio Hardware and Software

  • The radio uses software-defined radios and custom phased arrays, with a 32-antenna 4x8 array configuration and dual polarization.The arrays operate at 28 GHz and can switch beams in under 6 μs; FPGA-controlled registers load phase and gain settings.Source: §4.1 · pp. 8–9
  • The base-station radio uses an Ettus N310 USRP, while compact PlutoSDR-based user units support multi-user experiments.The waveform uses a modified 5G NR beam-management protocol with a 12.5 μs OFDM symbol and repeated beam IDs in the data.Source: §4.1 · p. 9

4.2 Radar Hardware and Software

  • The radar is an INRAS RadarBook FMCW unit operating at 24 GHz, with a 16-antenna virtual array formed from 2 transmit and 8 receive antennas.Its stated range, velocity, and angular resolutions are 0.75 m, 0.75 m/s, and 7.16 degrees, respectively.Source: §4.2 · p. 9
  • Radio and radar timestamps support synchronization, while trigger chaining aligns beam training with transmission or reception.Because the units are about 15 cm apart, radar points are transformed into a global coordinate system located at the radio.Source: §4.2 · p. 9

5 Evaluation

CommRad is evaluated across indoor and outdoor scenarios with multiple users and mobility patterns, against a non-collaborative tracking baseline, a reactive baseline, and an Oracle. Results cover throughput, overhead, user and reflector tracking, and blockage adaptation.

5.1 Overall Throughput Improvement

  • 1000 Mbps median throughput for CommRad versus 400 Mbps for the non-collaborative baseline yields a 2.5x gain; the 20th-percentile throughput rises from 50 Mbps to 400 Mbps, an 8x gain.CommRad performs almost similarly to the Oracle; Figure 15 presents the end-to-end results.Source: §5.1 · p. 10, Figure 15 · p. 10
  • Constructive multi-beam throughput drops by half with blockage, compared with an almost 10x drop for single-beam, and mean throughput increases by about 100 Mbps without blockers.The constructive scheme uses direct and reflected beams together.Source: §5.1 · p. 10

5.2 Overhead Reduction

  • CommRad reduces overhead to 0.5% with radar-based beam management at 0.5-second periodicity, and at 2 seconds the overhead is further reduced by 0.25%.Radar can indicate a new object's arrival and angle, allowing the gNB to scan only a small subset of angles.Source: §5.2 · p. 11

5.3 Angle Tracking Accuracy

  • 4 degrees median angular error for CommRad compares with 8 degrees for the baseline; at the 90th percentile, CommRad's 10-degree error is below the baseline's more than 40 degrees.Around 2 seconds, error is around 7 degrees, within the radio's 10-degree beamwidth; four-user overlap also causes the non-collaborative tracker to lose user identity.Source: §5.3 · p. 11, Figure 17 · p. 11

5.4 User/Reflector Tracking Accuracy

  • Radar's reflected-path angle tracking matches the Oracle with high accuracy, identifying a reflector path at 2 seconds and its disappearance at 7 seconds.The paper attributes this to context about reflector location, orientation, and size; Figure 18 compares tracking with the Oracle.Source: §5.4 · p. 12, Figure 18 · p. 12

5.5 Blockage Prediction and Mitigation

  • CommRad switches to reflected paths at blockage events, maintaining high RSS while a direct-path baseline without adaptation loses more than 10 dB.The trace shows beam-angle changes at 3 and 5 seconds.Source: §5.5 · p. 12

6 Discussion and Future work

CommRad’s discussion identifies its 24 GHz ISM radar as the current sensing setup and considers broader sensor fusion. Its beam-tracking design also assumes a quasi-omnidirectional user-side beam.

6 Discussion and Future work

  • CommRad deploys a 24 GHz radar in the ISM band for context-aware beam management.The discussion considers fusion with other sensing modalities as a future direction.Source: §6 · p. 12
  • Contextual awareness could be fused with 60 or 77 GHz radars, cameras, or lidars, and the framework is described as generalizable to other sensing modalities.Source: §6 · p. 12
  • Base-station beam tracking assumes a quasi-omnidirectional user-side beam; IMU-based user-beam adaptation is suggested as a complementary technique.Source: §6 · p. 12

7 Conclusion

The conclusion presents radar-radio collaboration as a way to improve mobile mmWave links. It reports improved link quality even under blockages and mobility-related beam misalignment.

7 Conclusion

  • Radar and radio combine complementary monostatic and bistatic sensing information to improve mobile mmWave link quality under blockages and mobility-related beam misalignment.Source: §7 · p. 12
Loading 2407.08817v1…