Sensing Assisted Channel Estimation for ISAC

Introduction

Channel estimation is fundamental to any wireless communication system. In OFDM-based 5G NR and beyond, channel estimation relies on reference signals (CSI-RS, DMRS) transmitted at known time-frequency positions. The UE estimates the channel at these pilot locations and interpolates across the resource grid. However, the quality of channel estimation is limited by the pilot density, SNR, and the channel’s delay-Doppler characteristics.

Sensing-assisted channel estimation exploits the fact that radar sensing and communication channels share the same physical propagation environment. The sensing subsystem can extract physical channel parameters — such as path delays, Doppler shifts, angles of arrival/departure, and path gains — from echo signals. These parameters can be used to construct a parametric channel model that complements or replaces pilot-based estimation.

 Conventional vs. Sensing-Assisted Channel Estimation
┌─────────────────────────────────────────────────────────────────────┐
│                                                                     │
│   Conventional:                                                     │
│   ┌─────────┐    ┌───────────────┐    ┌────────────────┐            │
│   │ CSI-RS / │───▶│ LS/MMSE       │───▶│ Interpolated   │            │
│   │ DMRS     │    │ Estimation    │    │ Channel Matrix │            │
│   └─────────┘    └───────────────┘    └────────────────┘            │
│                                                                     │
│   Sensing-Assisted:                                                 │
│   ┌─────────┐    ┌───────────────┐    ┌────────────────┐            │
│   │ Sensing  │───▶│ Parameter     │───▶│ Parametric     │            │
│   │ Echo     │    │ Extraction    │    │ Channel (τ,ν,θ)│            │
│   └─────────┘    └───────┬───────┘    └───────┬────────┘            │
│                          │                    │                     │
│   ┌─────────┐    ┌───────▼───────┐    ┌───────▼────────┐            │
│   │ CSI-RS / │───▶│ Aided/Hybrid  │───▶│ Enhanced       │            │
│   │ DMRS     │    │ Estimation    │    │ Channel Matrix │            │
│   └─────────┘    └───────────────┘    └────────────────┘            │
│                                                                     │
└─────────────────────────────────────────────────────────────────────┘

Key Concepts

Parametric Channel Construction from Sensing

The wireless channel between a transmitter and receiver can be described as a sum of multipath components (MPCs), each characterized by:

Delay \(\\tau_l\): propagation delay of the l-th path
Doppler shift \(\\nu_l\): Doppler frequency due to relative motion
Angle of departure \(\\theta_{\\text{AoD},l}\) and angle of arrival \(\\theta_{\\text{AoA},l}\)
Complex gain \(\\alpha_l\): amplitude and phase of the path

Sensing signal processing (e.g., 2D-FFT, MUSIC, ESPRIT) can estimate \((\\tau_l, \\nu_l, \\theta_l)\) for the dominant paths. Given these parameters, a parametric channel can be constructed as:

\[\begin{split}\\mathbf{H}(f, t) = \\sum_{l=1}^{L} \\alpha_l \\, \\mathbf{a}_{\\text{Rx}}(\\theta_{\\text{AoA},l}) \\, \\mathbf{a}_{\\text{Tx}}^H(\\theta_{\\text{AoD},l}) \\, e^{-j2\\pi f \\tau_l} \\, e^{j2\\pi \\nu_l t}\end{split}\]

where \(\\mathbf{a}_{\\text{Tx}}\) and \(\\mathbf{a}_{\\text{Rx}}\) are the transmit and receive array steering vectors.

Hybrid Estimation: Combining Sensing and Pilots

Rather than replacing pilot-based estimation entirely, sensing information can be used as a prior to improve estimation accuracy:

Sparse channel estimation: Sensing identifies the delay-Doppler support of the channel, enabling compressed sensing techniques with fewer pilots.
Pilot density reduction: If the parametric channel from sensing is sufficiently accurate, the pilot density can be reduced, freeing resources for data transmission.
Extrapolation in time: Sensing-derived Doppler information enables prediction of the channel into the future, reducing feedback latency effects.

 Hybrid Estimation Pipeline
┌───────────────────────────────────────────────────────────────────┐
│                                                                   │
│   Sensing Path:                                                   │
│   ┌────────┐   ┌──────────┐   ┌─────────────┐                    │
│   │ Echo   │──▶│ Delay-   │──▶│ Parametric  │──┐                 │
│   │ Signal │   │ Doppler  │   │ Channel     │  │                 │
│   └────────┘   │ Analysis │   │ H_sensing   │  │                 │
│                └──────────┘   └─────────────┘  │                 │
│                                                │  ┌────────────┐ │
│                                                ├─▶│  Fusion /  │ │
│                                                │  │  Weighted  │ │
│   Pilot Path:                                  │  │  Combine   │ │
│   ┌────────┐   ┌──────────┐   ┌─────────────┐ │  └─────┬──────┘ │
│   │ CSI-RS │──▶│ LS/MMSE  │──▶│ Pilot-based │─┘        │        │
│   │ Pilots │   │ Estimator│   │ H_pilot     │           ▼        │
│   └────────┘   └──────────┘   └─────────────┘   ┌────────────┐   │
│                                                  │ H_combined │   │
│                                                  │ (Enhanced) │   │
│                                                  └────────────┘   │
│                                                                   │
└───────────────────────────────────────────────────────────────────┘

Benefits

Table 5 Benefits of Sensing-Assisted Channel Estimation
Benefit	Description
Reduced pilot overhead	Sensing provides prior knowledge of the channel support, enabling sparser pilot grids.
Improved accuracy in high mobility	Sensing-derived Doppler enables better tracking of fast-varying channels.
Channel prediction	The parametric model enables extrapolation beyond the measurement window.
Robustness at low SNR	Sensing priors regularize the estimation problem, improving performance at low pilot SNR.

Challenges

Sensing-communication channel mismatch: The monostatic sensing channel (gNB → target → gNB) differs from the communication channel (gNB → UE). Mapping between the two requires assumptions about the scattering environment.
Parameter association: Associating sensing-detected paths with communication channel paths is non-trivial, especially in dense multipath environments.
Computational complexity: Joint delay-Doppler-angle estimation using super-resolution algorithms (MUSIC, ESPRIT) is computationally expensive.
Calibration: Accurate parametric channel construction requires well-calibrated antenna arrays and timing synchronization between the sensing and communication subsystems.