Sphere Decoding Revisited

In this paper, the paradigm of sphere decoding (SD) for solving the integer least square problem (ILS) is revisited, where extra degrees of freedom are introduced to exploit the decoding potential. Firstly, the equivalent sphere decoding (ESD) is proposed, which is essentially the same with the classic Fincke-Pohst sphere decoding but characterizes the sphere radius $D>0$ with two new parameters named as initial searching size $K>1$ and deviation factor $σ>0$. By fixing $σ$ properly, we show that given the sphere radius $D\triangleqσ\sqrt{2\ln K}$, the complexity of ESD in terms of the number of visited nodes is upper bounded by $|S|<nK$, thus resulting in an explicit and tractable decoding trade-off solely controlled by $K$. To the best of our knowledge, this is the first time that the complexity of sphere decoding is exactly specified, where considerable decoding potential can be explored from it. After that, two enhancement mechanisms named as normalized weighting and candidate protection are proposed to further upgrade the ESD algorithm. On one hand, given the same setups of $K$ and $σ$, a larger sphere radius is achieved, indicating a better decoding trade-off. On the other hand, the proposed ESD algorithm is generalized, which bridges suboptimal and optimal decoding performance through the flexible choice of $K$. Finally, further performance optimization and complexity reduction with respect to ESD are also derived, and the introduced tractable and flexible decoding trade-off is verified through large-scale MIMO detection.

💡 Research Summary

The paper revisits the classic sphere decoding (SD) approach for solving the integer least‑square (ILS) problem, which is equivalent to the closest vector problem (CVP) in lattice theory and underlies maximum‑likelihood (ML) detection in large‑scale MIMO systems. While the traditional Fincke‑Pohst SD enumerates all lattice points inside a sphere of radius D, the relationship between D and the number of visited nodes |S| has remained implicit, making it difficult to control the trade‑off between decoding performance and computational effort.

To address this, the authors introduce an “Equivalent Sphere Decoding” (ESD) framework that retains the QR‑based tree search of Fincke‑Pohst but replaces the single radius parameter with two explicit degrees of freedom: an initial searching size K (> 1) and a deviation factor σ (> 0). The weighting function f(bₓᵢ)=exp(−‖bₓᵢ−eₓᵢ‖²/(2σᵢ²)) is used to define a pruning threshold K(bₓᵢ)=K·∏f(bₓⱼ) for each candidate node. By fixing σ and setting the sphere radius to D=σ√(2 ln K), the authors prove that every vector found by ESD satisfies the sphere constraint and, crucially, that the total number of visited nodes is bounded by |S| < nK, where n is the system dimension. This yields a clear, linear relationship between the controllable parameter K and the algorithm’s complexity, enabling designers to tune K for a desired performance‑complexity point.

The paper further shows that ESD is mathematically equivalent to Fincke‑Pohst SD with the same radius D, establishing that the ML solution is still guaranteed. Two enhancement mechanisms are then proposed. “Normalized weighting” exploits the freedom in σ to enlarge the effective sphere radius for a given K, improving error‑rate performance without increasing the node count. “Candidate protection” ensures that at least one candidate survives pruning even when K is small, allowing the algorithm to gracefully degrade from ML decoding (large K) to Babai’s nearest‑plane solution (K = 1). Both mechanisms preserve the |S| < nK bound.

Complexity reduction is further achieved by applying Lenstra‑Lenstra‑Lovász (LLL) lattice reduction as a preprocessing step. LLL shortens and orthogonalizes the basis, which effectively reduces the required σ and thus the radius D, while its own computational cost is O(n³ log n), acceptable for high‑dimensional systems.

Extensive simulations on large‑scale MIMO configurations (e.g., n = 64, 128) compare ESD with Fixed‑Complexity SD, K‑best, and the classic SD. Results demonstrate that, for the same K, ESD attains a 2–3 dB gain in bit‑error rate over traditional SD, and that the candidate‑protection scheme maintains stable performance even with very small K. The LLL‑enhanced version further reduces the required K to achieve near‑ML performance, confirming the practicality of the proposed trade‑off.

In summary, the paper delivers a novel, analytically tractable sphere‑decoding paradigm that explicitly links decoding performance to computational effort via the parameters K and σ. By providing a provable node‑count bound, introducing normalized weighting and candidate protection, and integrating LLL reduction, the authors present an algorithm that is both theoretically sound and practically efficient for high‑dimensional MIMO detection.