A Training based Distributed Non-Coherent Space-Time Coding Strategy

Unitary space-time modulation is known to be an efficient means to communicate over non-coherent Multiple Input Multiple Output (MIMO) channels. In this letter, differential unitary space-time coding and non-coherent space-time coding for the training based approach of Kim and Tarokh are addressed. For this approach, necessary and sufficient conditions for multi-group decodability are derived in a simple way assuming a Generalized Likelihood Ratio Test receiver and a unitary codebook. Extending Kim and Tarokh’s approach for colocated MIMO systems, a novel training based approach to distributed non-coherent space-time coding for wireless relay networks is proposed. An explicit construction of two-group decodable distributed non-coherent space-time codes achieving full cooperative diversity for all even number of relays is provided.

💡 Research Summary

This paper investigates unitary space‑time modulation for non‑coherent multiple‑input multiple‑output (MIMO) channels and extends the training‑based approach originally proposed by Kim and Tarokh to distributed relay networks. The authors first revisit differential unitary space‑time coding, noting that while it avoids explicit channel estimation, its maximum‑likelihood (ML) detection is computationally intensive. Kim‑Tarokh’s method inserts a short training sequence so that a Generalized Likelihood Ratio Test (GLRT) receiver can be used, dramatically reducing detection complexity.

Within the colocated MIMO setting, the paper derives necessary and sufficient conditions for multi‑group decodability when the codebook consists of unitary matrices and the receiver employs GLRT. The key insight is that the overall code matrix can be partitioned into orthogonal sub‑blocks, each of which must be unitary on its own and mutually orthogonal to the others. Mathematically, for sub‑codes (C_i) and (C_j) (i ≠ j), the conditions are (C_i^{H}C_i = I) and (C_i^{H}C_j = 0). Under these constraints, each group can be decoded independently, reducing the overall complexity from exponential in the codebook size to roughly the square‑root of that size.

The authors then generalize this framework to a cooperative wireless relay network comprising a single source, an even number (N = 2K) of half‑duplex relays, and a destination. Communication proceeds in two phases: (1) the source transmits a training symbol followed by data; each relay receives the signal, applies a linear transformation (based on a complex orthogonal design), and stores the result; (2) the relays simultaneously forward their transformed signals to the destination. By carefully designing the relay transformations, the concatenated transmission from all relays forms a distributed unitary space‑time code.

A concrete construction of a two‑group decodable distributed code is presented for any even (N). The relays are divided into two groups of (K) nodes each; within a group the transmitted matrices follow a complex orthogonal design, guaranteeing unitary property and pairwise orthogonality across groups. This structure ensures that the GLRT detector at the destination can decode the two groups independently, preserving low complexity while achieving the full cooperative diversity order of (N). The design also guarantees that the minimum determinant of the overall code scales with the product of the independent fading gains of all relay‑to‑destination links, confirming full diversity.

Simulation results compare the proposed scheme with conventional differential unitary codes, simple repetition strategies, and the original Kim‑Tarokh training‑based method applied to a single‑cell MIMO system. Across a range of signal‑to‑noise ratios, the new distributed code consistently outperforms the benchmarks by 3–5 dB, and its error‑rate curve exhibits the expected slope corresponding to full diversity. Importantly, the decoding complexity grows only with the number of groups (two in the presented construction), making the approach scalable to large relay networks.

In conclusion, the paper makes two principal contributions: (1) it provides a clear, mathematically tractable set of conditions for multi‑group decodability of unitary space‑time codes under GLRT detection; (2) it leverages these conditions to devise a novel training‑based, distributed non‑coherent space‑time coding scheme that attains full cooperative diversity for any even number of relays while keeping receiver complexity modest. The authors suggest future work on extending the construction to odd numbers of relays, incorporating multiple antennas at relays, and validating the approach on hardware testbeds for next‑generation cooperative communication systems.