Embedded Pilot Channel Estimation

• Embedded pilot channel estimation is a technique that embeds structured pilot symbols within data streams to enable real-time channel estimation and coherent data decoding.
• It optimizes spectral efficiency by reducing pilot overhead using impulse, spread, and index-modulated pilots while mitigating interference via guard zones and adaptive grids.
• These methods are applied in advanced systems like OTFS, FTN, AFDM, and MIMO, offering improved performance in high-mobility and highly dispersive channels.

Embedded pilot channel estimation refers to a family of channel identification and tracking methods wherein known reference (pilot) symbols are interleaved within data transmission blocks, allowing joint or sequential estimation of the channel impulse response and decoding of user data. These schemes are designed to manage spectral efficiency, pilot overhead, complexity, and robustness to time- and frequency-selectivity, especially in the context of modern communication paradigms such as OTFS, FTN, AFDM, and advanced MIMO waveforms (Keykhosravi et al., 2024, Wang et al., 2 Dec 2025, Yin et al., 2022). Embedded pilot frameworks exploit specific symbol arrangements, sequence design, and detection algorithms to achieve precise channel estimation even in highly dispersive or high-mobility channels while minimizing signaling overhead.

1. Principles and Rationale of Embedded Pilot Channel Estimation

Embedded pilot channel estimation is motivated by the need to estimate rapidly time-varying or frequency-selective fading channels with minimal resource overhead. Unlike isolated training (where pilots are sent in dedicated training-only slots), embedded schemes strategically insert pilot symbols or sequences amid user data—either as explicit symbols (impulses, guard zones, frequency tones) or more general structured signals (e.g., MLS, Zadoff-Chu, or index-modulated pilots) (Keykhosravi et al., 2024, Sun et al., 2024, Wang et al., 2 Dec 2025). This enables real-time or frame-wise acquisition of channel state information (CSI) that is contemporaneous with the data, critically supporting coherent detection in scenarios with limited coherence time or frequency.

These strategies are particularly effective in systems where pilot overhead must be tightly constrained. For example, in OTFS or AFDM, conventional pilot overhead can dominate for long channel delay spreads or many users; embedding and structure-aware allocation address this by providing localized or pilot-overlapping regions in the time-frequency or delay-Doppler plane (Raviteja et al., 2018, Yin et al., 2022).

2. Embedded Pilot Design: Placement, Types, and Protocols

Embedded pilot strategies vary substantially across waveform domains and application regimes, but all share three essential components: pilot symbol design, placement/allocation, and accompanying guard strategies.

Pilot Design and Sequence Types:

• Impulse Pilots: Single tone or DD-domain impulses, possibly with frequency/space/guard separation to minimize interference (Raviteja et al., 2018, Li et al., 2024).
• Spread Sequences: PN, Zadoff-Chu (ZC), and maximum-length sequences (MLS) are adopted for their sharp correlation and low sidelobe properties, enabling robust multi-user separation or pilot location identification (Sun et al., 2024, Wang et al., 2 Dec 2025).
• Index-modulated pilots: The position or structure of the pilot sequence itself conveys data (index modulation), increasing net rate (Keykhosravi et al., 2024).

Placement and Guard Strategies:

• Fixed vs. Adaptive Grids: Pilots are placed periodically, on a grid that aligns with expected delay or Doppler spread, or their locations are indexed to users or codewords (Keykhosravi et al., 2024, Yin et al., 2022).
• Guard Zoning: To isolate pilots from data-induced ISI/ICI, regimes such as delay/Doppler guard zones or zero padding are applied (Raviteja et al., 2018, Li et al., 2024, Aghda et al., 2023).
• Superimposed Pilots: Pilots are linearly superimposed on data without guard symbols, relying on orthogonality or iterative interference cancellation to separate signals (Zheng et al., 2024).

3. Channel Estimation Algorithms and Detection Procedures

The channel estimation process in embedded pilot schemes typically comprises pilot detection/matching and path parameter estimation:

• Pilot Position Recovery: For index-modulated or user-overlapped pilots, pilot location identification at the receiver can be accomplished via correlation maximization or by exploiting cyclic/orthogonality properties of the pilot sequences (Keykhosravi et al., 2024, Wang et al., 2 Dec 2025).
• Energy or Correlation Thresholding: Estimation often involves constructing correlation peaks between received samples and pilots (or their autocorrelation), followed by threshold-based detection of path presence (SPA/MPA in AFDM, OTFS EPA) (Raviteja et al., 2018, Yin et al., 2022, Sun et al., 2024).
• LS/LMMSE/Sparse Recovery: Given the pilot output blocks, LSSE/LMMSE solutions or OMP/sparse recovery techniques are employed as appropriate to the effective channel model and prior (Aghda et al., 2023, Zheng et al., 2024, Wang et al., 2 Dec 2025).
• Iterative Embedded Detector-Estimator: Iterative frameworks jointly cancel pilot-data interference and refine both channel and data estimates (e.g., in AFDM with superimposed pilots or OTFS split-pilot methods) (Zheng et al., 2024, Li et al., 2024).
• Specialized Operators: For phaseless schemes (OFDM), phase retrieval procedures using alternating projections or semidefinite relaxation replace classical LS channel estimation (Walk et al., 2015).

4. Overhead, Spectral Efficiency, and Performance Trade-Offs

Embedded pilot estimation methods are extensively evaluated across several axes: pilot overhead, spectral efficiency, estimation error (MSE/NMSE), BER, PAPR, and computational complexity.

Overhead and Efficiency Optimization:

• Guard-based approaches (embedded pilot with explicit guard zones) incur pilot+guard overhead proportional to path delay/Doppler spread (Raviteja et al., 2018, Li et al., 2024).
• Superimposed pilots, index-modulation, and cyclic shift-based embeddings achieve significant overhead reductions (up to 30–40%) by removing all or most guard zones, increasing SE (Keykhosravi et al., 2024, Wang et al., 2 Dec 2025, Zheng et al., 2024).

Performance Metrics:

• SE improvements for index-modulated FTN pilots can be as high as 45% over Nyquist reference, with up to 1–5% over traditional FTN pilots at similar channel estimation error (Keykhosravi et al., 2024).
• BER and NMSE are used as primary channel estimation quality indicators; several schemes achieve performance within 0.2–0.5 dB of ideal CSI (Sun et al., 2024, Yin et al., 2022, Li et al., 2024).
• Complexity is dominated by steps such as exhaustive pilot search (for location ID), matrix inversion, or grid-based sparse recovery, but recent methods ensure $O(N\log N)$ or lower computational profiles (Sun et al., 2024, Li et al., 2024).

Significant Trade-Offs:

• While reducing guards/superimposing pilots improves SE, it increases pilot-data interference and may require iterative refinement or increased pilot power (Zheng et al., 2024).
• In multi-user or multi-antenna settings, pilot orthogonality via ZC sequence cyclic shifts provides scalable overhead without sacrificing estimation accuracy, provided user-specific shifts are carefully assigned (Wang et al., 2 Dec 2025).

5. Applications Across Waveforms and Scenarios

Embedded pilot-based channel estimation techniques have broad applicability:

• OTFS: Embedded pilots (impulse/MLS) and reduced-guard/split-pilot mechanisms provide robust estimation in doubly dispersive, multiuser scenarios and are essential for high-mobility links (Raviteja et al., 2018, Sun et al., 2024, Li et al., 2024, Wang et al., 2 Dec 2025).
• FTN: Index-modulated pilots enable improved SE and robust pilot block detection even in doubly selective HF channels with stringent multipath constraints (Keykhosravi et al., 2024).
• AFDM: SPA/MPA and superimposed pilot regimes trade off between ultra-low overhead and guard-free transmission, supporting both single- and multiuser scenarios (Yin et al., 2022, Zheng et al., 2024).
• OFDM: Phaseless (magnitude-only) pilot protocols can decouple amplitude and phase knowledge, enabling phase resources to carry auxiliary information or reduce PAPR, critical for short-packet/low-latency links (Walk et al., 2015).
• Optical/MIMO/Multichannel: Optimized multidimensional pilot allocation for joint-channel phase-noise estimation and high-order modulation, with performance depending critically on cross-channel correlation and modulation format (Alfredsson et al., 2020).

6. Advanced Topics: Multi-User, MIMO, and Ambiguity Resolution

Recent advances extend embedded pilot estimation to multi-user and MIMO contexts:

• Cyclic Shift Embedded Pilot (CSEP) in MIMO-OTFS: Exploits cyclic shift orthogonality of ZC sequences to allow overlapping pilots for multiple users, tightly packing pilots in the delay axis and achieving over 30% pilot overhead reduction without significant inter-user pilot interference (Wang et al., 2 Dec 2025).
• Fractional Delay/Doppler Estimation: Off-grid and sparse recovery methods iteratively refine path delay/Doppler to sub-bin accuracy (within 0.02 grid steps), overcoming the model mismatch inherent in on-grid approaches (Wang et al., 2 Dec 2025).
• Doppler Ambiguity in High-Velocity OTFS: Embedded pilots with guard zones, together with phase-difference ambiguity resolution, enable accurate channel estimation when Doppler supports exceed the fundamental lattice, avoiding model-induced BER/NMSE floors (Chen et al., 28 Jan 2026).

7. Implementation Complexity and Design Guidelines

Effective deployment of embedded pilot channel estimation schemes requires careful consideration of complexity, resource allocation, and system parameters:

• Computational Load: Most schemes achieve $O(N\log N)$ per-frame complexity via FFT-based transforms and sparse updates (Sun et al., 2024, Aghda et al., 2023).
• Pilot Power Allocation: Safeguarding estimation quality in superimposed pilot designs demands substantially higher pilot power than data (e.g., SNR$_p \approx 50$ dB) to mitigate pilot-data interference, with path thresholding set adaptively as a function of noise (Zheng et al., 2024).
• Guard/Pilot Sizing: Guard sizes are matched to maximum anticipated channel spread; insufficient guards lead to pilot-data leakage and estimation error floors (Raviteja et al., 2018, Li et al., 2024), while excess guards waste capacity.
• Threshold Setting: Detection and estimation step thresholds must be tuned by simulation or analytical bounds (e.g., $3\sigma$ for AWGN) to balance false alarms and missed taps (Yin et al., 2022).
• Pilot Position Optimization: In multidimensional (e.g., optical) environments, pilot positions are jointly optimized via structured/unstructured design to minimize estimator MSE, substantially improving achievable information rate for high-order modulation and high channel correlation (Alfredsson et al., 2020).

In summary, embedded pilot channel estimation forms the backbone of contemporary robust operation in highly dynamic, spectrally-constrained communication systems, supporting ultra-reliable, high-mobility applications across a spectrum of physical layer designs and network topologies (Keykhosravi et al., 2024, Wang et al., 2 Dec 2025, Yin et al., 2022, Raviteja et al., 2018).