Performance Comparison of Joint Delay-Doppler Estimation Algorithms

TL;DR

This paper tackles the challenge of real-time joint delay-Doppler estimation for ISAC by comparing three representative algorithms—RIMAX (model-based), DeepEst (CNN-based), and OS-CFAR (CFAR-based)—using publicly available sub-6 GHz bistatic channel data with two analytically characterized targets. Performance is quantified via target detection probability $P_D$, RMSEs of delay and Doppler, and runtime, with ground truth derived from the measurement geometry. In bistatic configurations, all three approaches exhibit similar estimation capabilities with $P_D$ reaching up to about 0.8, while forward and backward scattering dominated by a strong LoS component reduces detection to near zero. RIMAX and DeepEst deliver higher-resolution estimates at the cost of longer runtimes, whereas OS-CFAR provides a faster baseline that benefits from background subtraction and interpolation; DeepEst can misclassify static paths without proper preprocessing. Collectively, the results inform the design of real-time ISAC sensing pipelines and highlight the trade-offs between accuracy and computation in different scattering regimes, suggesting that high-resolution methods perform best in cluttered environments while CFAR-based methods can be competitive with appropriate preprocessing.

Abstract

Integrated sensing and communications (ISAC), radar, and beamforming require real-time, high-resolution estimation algorithms to determine delay-Doppler values of specular paths within the wireless propagation channel. Our contribution is the measurement-based performance comparison of the delay-Doppler estimation between three different algorithms, comprising maximum likelihood (ML), convolutional neural network (CNN), and constant false alarm rate (CFAR) approaches. We apply these algorithms to publicly available channel data which includes two spherical targets with analytically describable delay-Doppler parameters. The comparison of the three algorithms features the target detection rate, root mean squared errors (RMSEs) of the delay-Doppler estimates, and a runtime analysis. Notably, all three algorithms demonstrate similar parameter estimation capabilities in bi-static scenarios, achieving target detection probabilities of up to 80%. Conversely, forward and backward scattering conditions pose a problem to the estimation due to strong line-of-sight (LoS) contribution, reducing the corresponding detection probability down to 0%.

Figures (5)

• Figure 1: Bistatic Measurement Setup---Two metallic spheres (red) are mounted on a rotating beam. The bistatic measurement angle $\delta$ between and creates either backward, forward, or bistatic scattering scenarios. The length of the beam is 3 and the distances of and to the turntable center are 3.48 and 2.86, respectively.
• Figure 2: Empirical Target Detection Probability Compared to the Strength---While the solid lines depict this metric for an identification boundary of $\epsilon = 0.5$, the dashed lines represent $\epsilon = 0.25$.
• Figure 3: Delay-Doppler Estimation Results for a Bistatic Observation Angle of $\delta = \qty{20}{\degree}$---All plots include the results for one full rotation of the target emulator. DeepEst estimates a large number of static paths due to the missing background subtraction.
• Figure 4: Delay and Doppler --- (blue), RIMAX (green), and DeepEst (orange) exhibit similar both in magnitude and progression. Due to an incorrect classification of static paths, the latter algorithm yields increased errors in forward scattering scenarios.
• Figure :