Indoor Occupancy Classification using a Compact Hybrid Quantum-Classical Model Enabled by a Physics-Informed Radar Digital Twin

TL;DR

This work tackles privacy-preserving indoor occupancy sensing using mmWave radar, proposing a compact two-qubit Hybrid Quantum Neural Network (HQNN) and benchmarking it against CNN baselines. A physics-informed 60 GHz digital twin and real radar data are used to compare performance, revealing that the HQNN achieves high accuracy with ~66k parameters and demonstrates a distinct noise-sensitivity profile. Ablation experiments show that replacing the quantum layer with parameter-matched classical heads collapses performance, indicating the PQC provides genuine representational capacity beyond a simple bottleneck. The results establish a baseline for hybrid quantum models in privacy-preserving radar sensing, while highlighting the need for hardware validation and digital-twin calibration to translate these gains to real-world deployments.

Abstract

Indoor occupancy classification enables privacy-preserving monitoring in settings such as remote elder care, where presence information helps triage alarms without cameras or wearables. Radar suits this role by sensing motion through occlusions and in darkness. Modern deep-learning pipelines are the standard for interpreting radar returns effectively; however, they are often parameter-heavy and sensitive at low signal-to-noise ratios (SNR), motivating compact alternatives like Hybrid Quantum Neural Networks (HQNNs). A two-qubit HQNN is benchmarked against convolutional neural networks (CNNs) using a physics-informed 60GHz digital twin and real radar measurements under matched training protocols. In clean conditions, the HQNN achieves high accuracy (99.7% synthetic; 97.0% real) with up to 170x fewer parameters (0.066M). Its parameter efficiency is shown to be structural, as an ablation of the parameterized quantum circuit (PQC) causes sharp performance drops on real data (to 68.5% and 31.5% for the control heads). A domain-dependent sensitivity emerges under additive-noise evaluation, where the HQNN begins recovery earlier in synthetic data while CNNs recover more steeply and peak higher on real measurements. In label-fraction ablations, CNNs prove more sample-efficient on real Range-Doppler Maps (RDMs), with the performance gap being most pronounced (at 50% labels, BA 0.89-0.99 vs. HQNN 0.75). On synthetic data, this gap narrows significantly, largely vanishing by the 50% label mark. Overall, the HQNN's value lies in parameter efficiency and a compact inductive bias that shapes its distinct sensitivity profile; this work establishes a rigorous baseline for hybrid quantum models in privacy-preserving radar occupancy sensing.

Figures (18)

• Figure 1: Configuration modules of the digital-twin pipeline for FMCW radar simulation. (a) Actor configuration: animated human models are imported using DAE files with parameterized motion, scale, and velocity. (b) Room configuration: environment geometry is loaded from STL models with material assignments defining electromagnetic properties such as reflectivity and transparency. (c) Sensor configuration: FMCW radar front-end, antenna placement, and waveform settings are specified to replicate the 60 GHz device in simulation. Together, these modules enable a reproducible and physically grounded digital-twin setup that captures human motion, clutter, and multipath propagation.
• Figure 2: Digital-twin pipeline stages following scene configuration. (a) The simulation environment, defined programmatically, is executed and exported to a full-wave/high-frequency solver with assigned materials and radar parameters. The SBR method models multipath propagation, occlusions, and geometry-dependent interactions. (b) Radar signal processing converts the simulated baseband signals into RDMs, which form the dataset for training and evaluation of machine learning models.
• Figure 3: Schematic layout of the simulated corridor environment (12 m $\times$ 2 m) showing radar placement, representative multipath, and clutter/noise regions.
• Figure 4: Schematic layout of the simulated room environment (8 m $\times$ 6 m) including furniture (sofa, chair) and representative human activities.
• Figure 5: Simulated indoor environments and corresponding range--Doppler maps (RDMs). (a) Room with furniture and two occupants; (b) digital environment with furniture only; (c) corridor with one walking occupant. (d)--(f) RDMs for (a)--(c). Each RDM represents a single snapshot in time rather than the full scenario duration.