Adaptive Internal Calibration for Temperature-Robust mmWave FMCW Radars

TL;DR

This work tackles temperature-induced drift in mmWave FMCW radars deployed in dense networks by introducing an internal, privacy-preserving calibration framework that relies solely on onboard sensor data. The method builds a temperature-to-amplitude mapping $\mathcal{M}_{a,b}$ and applies a multiplicative correction $\tilde{A}_{f,a,b} = \bar{A}_{f,a,b} \cdot \frac{\mathcal{M}_{a}(T_{\text{REF}})}{\mathcal{M}_{a}(T_f)}$ to compensate gain changes, with online training and inference phases to adapt to operating conditions. Key contributions include a structured preprocessing pipeline (Bias Removal, FFT, Real-valued Frequency Selection, Chirp Averaging), per-antenna calibration, and experimental validation across two settings showing reduced temperature–amplitude coupling and robust performance with minimal computational load. The framework supports privacy-sensitive, scalable deployment in dense networks and has potential applications in healthcare, infrastructure monitoring, and automotive sensing, especially within a temperature range of $30$–$45$ °C. Future work may explore nonlinear models, target-temperature estimation, environmental factors, multi-radar scenarios, and efficiency optimizations for edge devices.

Abstract

We present a novel internal calibration framework for Millimeter- Wave (mmWave) Frequency-Modulated Continuous-Wave (FMCW) radars to ensure robust performance under internal temperature variations, tailored for deployment in dense wireless networks. Our approach mitigates the impact of temperature-induced drifts in radar hardware, enhancing reliability. We propose a temperature compensation model that leverages internal sensor data and signal processing techniques to maintain measurement accuracy. Experimental results demonstrate improved robustness across a range of internal temperature conditions, with minimal computational overhead, ensuring scalability in dense network environments. The framework also incorporates ethical design principles, avoiding reliance on sensitive external data. The proposed scheme reduces the Pearson correlation between the amplitude of the Intermediate Frequency (IF) signal and internal temperature drift up to 84%, significantly mitigating the temperature drift.

Figures (3)

• Figure 1: Block diagram of the radar system architecture, detailing the signal processing pipeline for temperature-based amplitude calibration and data analysis. The diagram employs color-coding to distinguish functional components: green represents radar signal acquisition and preprocessing, pink dashed line in the middle denotes -based frequency-domain analysis, and red indicates internal temperature measurement. Received signals from multiple antennas, denoted by index $k$, are processed independently to capture amplitude displacements, enabling per-antenna calibration to mitigate temperature-induced variations in signal amplitude for enhanced measurement accuracy.
• Figure 2: Experimental setup: A) Infineon BGT60TR13C mmWave radar, B) Thermometer on metallic target, C) Metallic target, D) Keysight temperature reading device, E) Temperature control device.
• Figure 3: Comparison of experimental results across two settings. (a) for training/testing phases and for varying receive antennas, and internal radar temperature in the first setting. (b) for training/testing phases and for varying receive antennas, and internal radar temperature in the second setting. (c) Linear regression models for predicting amplitude based on internal radar temperature across different antennas in both settings, distinguished by color.