Multi-site Radar Systems for High-Precision Indoor Positioning and Tracking

Abstract: This paper introduces a high-precision indoor positioning and tracking method that utilizes multi-site single-input single-output (SISO) radar systems. We propose a novel velocity synthesis-assisted (VSA) localization algorithm that iteratively refines target position estimates within range bins by fusing radial velocity measurements from multiple radars. This approach ensures enhanced accuracy in both velocity and position estimation. Moreover, the inherent geometric constraints introduced by velocity synthesis enable the proposed algorithm to remain robust under low signal-to-noise ratio (SNR), severe multipath propagation, and large synchronization latency. Notably, our method eliminates the use of multiple-input-multiple-output (MIMO) configurations and stringent phase synchronization requirements, substantially reducing hardware complexity while maintaining high positioning accuracy. We define standardized reference trajectories to facilitate a comprehensive and reproducible performance evaluation. Extensive simulations and experimental validations demonstrate that our multi-site radar systems achieve centimeter-level tracking accuracy for human subjects, outperforming existing methods in complex trajectory tracking.

• The paper presents a novel velocity synthesis-assisted (VSA) fusion algorithm that integrates Doppler-derived velocities, achieving up to 48% improvement in tracking accuracy.
• It leverages coordinated commercial SISO FMCW radars to enable high-precision localization with a mean RMSE as low as 6 cm under challenging indoor conditions.
• Experimental and simulation results validate the system’s robustness in multi-target tracking and its resilience to multipath, low SNR, and synchronization errors.

Multi-site Radar Systems for High-Precision Indoor Positioning and Tracking

Introduction and Motivation

The paper "Multi-site Radar Systems for High-Precision Indoor Positioning and Tracking" (2604.15688) addresses the longstanding challenge of accurate, tag-free human tracking in indoor environments, proposing a robust, low-cost methodology based on multi-site single-input single-output (SISO) frequency-modulated continuous-wave (FMCW) radars. Traditional solutions frequently rely on vision or radio-frequency identification technologies with significant limitations—privacy concerns, tag requirements, or complex multi-input multi-output (MIMO) radar arrangements. The proposed approach eliminates the need for expensive MIMO transceivers and strict phase-level hardware synchronization, instead leveraging coordinated, commercially available SISO modules. The central contribution is a velocity synthesis-assisted (VSA) fusion algorithm, which exploits Doppler-derived radial velocities as geometric constraints, enabling highly accurate, low-latency multi-perspective localization and trajectory estimation even under severe multipath, low-SNR, and non-ideal synchronization conditions.

Review of State-of-the-Art and Problem Definition

Conventional indoor positioning comprises either tag-based approaches or passive methods, the latter including both vision and various RF modalities (WiFi, UWB, mmWave radar). Vision-based systems provide high accuracy but suffer from privacy and lighting/environmental concerns. RF-based schemes—specifically mmWave radars—offer high resolution and privacy benefits but are typically hampered by hardware complexity or the need for statistical learning-based data fusion prone to generalization errors.

Single-radar MIMO configurations are limited by aperture and practical array size (Figure 1), resulting in restricted angular resolution and poor scalability. Multi-site radar systems, while promising, introduce synchronization and data fusion complexity (Figure 2). Existing data fusion techniques can be categorized as direct-machine learning or geometry-based. Learning-based fusers are data-hungry and non-interpretable; geometry-based approaches, such as trilateration from distance-only measurements, are prone to ambiguity, high GDOP, and are sensitive to environmental non-idealities (Figure 3).

Figure 1: Indoor high-precision target single positioning alternatives, comparing MIMO single-site and SISO multi-site architectures.

Figure 2: Synchronization strategies in multi-site radar configurations, illustrating hardware/phase-based (a) and non-coherent/software-based NTP-triggered approaches (b).

The solution space therefore requires a method that—without recourse to dense arrays or intricate cabling—achieves high spatial resolution, robust tracking, and cost-effective deployment using commercial hardware.

Proposed System and VSA Algorithm

The authors introduce a distributed, non-coherent, NTP-triggered pair of SISO mmWave radars—each deploying FMCW waveforms and leveraging Doppler-derived radial velocities alongside range estimates. The principal innovation is the VSA fusion algorithm, which, through a discrete-grid search, fuses instantaneous Doppler (velocity) information from each node to resolve position and motion states with significantly increased geometric constraints. The VSA method proceeds as follows (Figure 4):

• Spatial Candidate Generation: Defines a grid within the overlap of per-radar range bins.
• Multi-Site Velocity Synthesis: For each grid point, the algorithm synthesizes candidate full velocity vectors using pairs of radial velocities from distinct radar nodes, applying a closed-form relation derived from LoS geometry.
• Spatiotemporal Consistency Enforcement: Candidates are accepted if their synthesized velocities exhibit magnitude and direction consistency across snapshots, as dictated by the local constant-velocity assumption.
• State Estimation Output: The optimal spatiotemporally consistent (position, velocity) pair is provided to the subsequent EKF track filter.

  Figure 4: The architecture of the proposed Velocity Synthesis-Assisted (VSA) method, showing geometric fusion via Doppler constraints.

Crucially, the VSA acts as a pre-filtering mechanism, enforcing physical constraints before feeding measurements to the dynamic tracker, resulting in marked improvements during rapid maneuvers or under non-ideal signal conditions.

Numerical Results and Simulation Analysis

Monte Carlo simulations benchmark VSA against conventional trilateration and EKF-based tracking. In single-target localization, VSA achieves a mean RMSE of 6 cm, a 48% improvement over conventional methods (Figure 5). Robustness analyses demonstrate persistent advantages under:

• Low SNR: At 0 dB SNR, VSA yields a position RMSE of 14 cm vs. 25 cm for EKF-only (Figure 6a).
• Multipath/Outlier Proportion: With 20% outlier measurement rate, VSA errors remain at the centimeter-level, while EKF errors accumulate linearly (Figure 6b).
• Synchronization Latency: With up to 50 ms inter-radar alignment error, VSA maintains low RMSE, validating tolerance to NTP synchronization errors (Figure 6c).

  Figure 5: Monte-Carlo simulations—(a) spatial distribution of VSA/range-only estimates; (b) histogram of positioning errors across trials.

  

  Figure 6: Monte-Carlo trajectory tracking robustness—(a) SNR sweep; (b) multipath/interference; (c) inter-radar latency; (d) pulses-in-frame; (e) grid granularity; (f) system throughput versus computational load.

Tracking performance was further interrogated via a suite of complex trajectories (rhombus, circular, star-shaped). VSA consistently outperforms baseline EKF, particularly on highly nonlinear/star paths (Figure 7), and suppresses sharp error spikes that coincide with trajectory discontinuities. Absolute RMSE is reduced from 28 cm to 13 cm on the star-shaped trajectory—a >50% error reduction—and the distribution of instantaneous error is substantially more concentrated.

Figure 7: Tracking results for (a) rhombus, (b) circular, and (c) star-shaped standard trajectories highlighting VSA advantages.

Experimental Verification

Physical deployment in a 9m×9m space using COTS AWR6843ISK radars validated these findings. The system accurately tracked both single and dual human targets. Compared to direct point-cloud fusion or state-of-the-art fusion algorithms, the VSA system delivered:

• Reduced RMSE: e.g., for rhombus trajectory, 10.5 cm (VSA) vs. 20.4 cm (point cloud) or 16.5 cm (SAV-PF).
• Complex Scenarios: For star-shaped motion, errors were consistently lower (16.3 cm vs. 22.5 cm).
• Multi-target Capability: Demonstrated error of 11.4 cm in face-to-face dual tracking.

  Figure 8: Physical experiment setup for single and multi-target tracking with SISO radars.

  

  Figure 9: Experimental trajectory tracking results comparing single MIMO and multi-site SISO systems.

  

  Figure 10: Empirical tracking performance for standard trajectories using the proposed system.

Direct cross-comparison with published systems (see Table IV in paper) demonstrates that the multi-site SISO-VSA approach attains comparable or superior accuracy to MIMO or IR-UWB based platforms, without recourse to expensive or onerous synchronization hardware.

Implications, Limitations, and Future Directions

The findings show that multi-site SISO radar systems, coupled with rigorous Doppler/velocity-informed fusion, offer a scalable and privacy-preserving alternative to camera-based or massive-MIMO indoor positioning. The VSA algorithm's resilience to multipath, SNR degradation, and synchronization error enables practical deployment in consumer-grade environments.

Practically, this opens the field to widespread smart-home, healthcare, and industry applications where low-cost, passive, and privacy-respecting real-time tracking is required. Theoretical implications revolve around the efficacy of physical-constraint-informed fusion layers as robust pre-filters for dynamic state estimation, especially under high maneuverability and ambiguous measurement conditions. The method’s grid-based search, while optimal for accuracy, comes with moderate increases in computational demand; real-time application is achieved but would benefit from parallelization or approximate search strategies.

For future research, comprehensive multi-target separation in highly cluttered environments remains an open problem. Integrating advanced motion models, real-time calibration, or unsupervised learning for dynamic thresholding and anomaly rejection may further enhance reliability.

Conclusion

This work rigorously establishes velocity synthesis-assisted multi-site SISO radar as a strong candidate for high-precision, cost-effective, and robust indoor positioning. Its physically constrained grid fusion architecture demonstrably outperforms both classic range-based and particle filter methods in simulation and experiment, with performance on par with or exceeding state-of-the-art MIMO/IR-UWB platforms. The demonstrated resilience to multiplicity of environmental and hardware artifacts underlines its suitability for scalable deployment. Future work should address computational scaling for multi-target tracking and deeper integration with environmental learning for dynamic environments.