Joint Phase Tracking and Channel Decoding for OFDM Physical-Layer Network Coding

Taotao Wang, Soung Chang LiewarXiv 2013 · OpenReview

TL;DR

This paper tackles the challenge of maintaining accurate phase synchronization in OFDM-based Physical-Layer Network Coding (PNC) systems, which are crucial for efficient wireless relaying. They propose a novel iterative approach that jointly optimizes phase tracking and channel decoding, leading to significant performance improvements over traditional methods.

What Problem Does This Paper Address?

In wireless networks, especially those employing relay nodes to extend coverage and improve reliability, efficient communication is key. Physical-Layer Network Coding (PNC) offers a way to boost throughput in two-way relay channels (TWRC) by allowing nodes to transmit simultaneously and the relay to combine their signals. Orthogonal Frequency Division Multiplexing (OFDM) is a popular modulation scheme for PNC due to its resilience to multipath fading and its ability to relax strict time synchronization requirements.

However, OFDM systems are sensitive to carrier frequency offsets (CFOs), which cause phase drifts. In traditional OFDM systems, pilot tones are used to track these phase drifts. In OFDM PNC, multiple nodes transmit simultaneously, meaning pilot tones must be shared, reducing the number available to each node. This makes phase tracking more difficult. Furthermore, the superimposed signals at the relay receiver create a denser constellation, complicating channel decoding. Ignoring these phase drifts leads to performance degradation and decoding failures. This paper addresses the problem of jointly performing phase tracking and channel decoding in OFDM PNC systems to overcome these challenges.

Methods

The core contribution of this paper is the development of a joint Expectation-Maximization (EM) and Belief Propagation (BP) algorithm (EM-BP) for OFDM PNC. This approach leverages an iterative process:

  • Belief Propagation (BP) for Channel Decoding: The BP algorithm is used to decode the channel, utilizing the phase-adjusted Channel State Information (CSI) estimated by the EM algorithm. It operates on a factor graph that models the constraints of virtual channel encoding for PNC.
  • Expectation-Maximization (EM) for Phase Tracking: The EM algorithm is employed to estimate the phase drifts. Crucially, it uses not only the pilot symbols but also the soft information about the data symbols provided by the BP algorithm. This is the key innovation – using data to improve phase tracking, which in turn improves data decoding.

The EM algorithm itself has two steps:

  • E-Step: Computes the Q function, which is a lower bound on the log-likelihood of the observed data given the current phase estimates.
  • M-Step: Maximizes the Q function to update the phase estimates. To efficiently solve this maximization, the authors employ a particle-filtering type method, which adaptively refines a set of candidate phase estimates.

The EM and BP algorithms are iterated, allowing them to mutually improve each other's performance. The process is initialized using a least-square (LS) estimation of phases from pilot tones.

Datasets

Not discussed in the paper. The evaluation is based on simulations and experiments using a prototype OFDM PNC system.

Key Insights and Ideas

  • Joint Optimization is Key: The most significant insight is that phase tracking and channel decoding are not independent problems in OFDM PNC. By jointly optimizing them iteratively, performance gains are achieved that are not possible when treating them separately.
  • Leveraging Data for Phase Tracking: The EM algorithm's ability to incorporate soft information from data symbols (provided by BP) for phase tracking is a crucial improvement over traditional pilot-only methods. This is particularly effective when pilot tones are scarce.
  • Particle Filtering for Efficient M-Step: The use of a particle-filtering approach in the M-step of the EM algorithm provides a practical and effective way to find the optimal phase estimates, especially in a complex, multi-dimensional search space.
  • Robustness to Fading: The EM-BP method shows improved robustness to deep fading on pilot tones, as it can rely on data symbol information when pilots are unreliable.

Key Results

  • Significant BER Gains: The proposed EM-BP algorithm achieves substantial Bit Error Rate (BER) improvements compared to traditional pilot-based phase tracking and channel decoding.
  • 2-3 dB Gain: Simulations and experiments demonstrate a gain of approximately 2 dB after the first iteration of the EM-BP algorithm and around 3 dB gain after convergence.
  • Improved Phase Estimation Accuracy: The Mean Square Error (MSE) of phase estimation is significantly reduced by the EM-BP method, indicating more accurate tracking.
  • Performance Across Channel Models: The gains are observed in both flat and frequency-selective Rayleigh fading channels, with similar trends.

Limitations and Future Work

  • Computational Complexity: While the paper uses a particle-filtering method for efficiency, iterative EM-BP algorithms can still be computationally intensive. The paper mentions "decoupled computation" to simplify the Q function, but the overall complexity is not explicitly detailed as a primary focus for optimization.
  • Channel State Information: The simulations assume perfect knowledge of the channel state information (CSI) at the relay. In real-world scenarios, CSI estimation itself can be imperfect, which might affect the performance of the proposed method.
  • Future Work: The paper implicitly suggests further exploration of optimizing the EM-BP framework, potentially by investigating different initialization strategies or more advanced particle filtering techniques. The impact of imperfect CSI could also be a direction for future research.

Summary updated

BibTeX
@misc{wang2013joint,
  author = {Taotao Wang and Soung Chang Liew},
  title = {{Joint Phase Tracking and Channel Decoding for OFDM Physical-Layer Network Coding}},
  year = {2013},
  url = {https://openreview.net/forum?id=X2YOVVlH1J},
}
Figure