Construction and Covering Properties of Constant-Dimension Codes

Constant-dimension codes (CDCs) have been investigated for noncoherent error correction in random network coding. The maximum cardinality of CDCs with given minimum distance and how to construct optimal CDCs are both open problems, although CDCs obtained by lifting Gabidulin codes, referred to as KK codes, are nearly optimal. In this paper, we first construct a new class of CDCs based on KK codes, referred to as augmented KK codes, whose cardinalities are greater than previously proposed CDCs. We then propose a low-complexity decoding algorithm for our augmented KK codes using that for KK codes. Our decoding algorithm corrects more errors than a bounded subspace distance decoder by taking advantage of the structure of our augmented KK codes. In the rest of the paper we investigate the covering properties of CDCs. We first derive bounds on the minimum cardinality of a CDC with a given covering radius and then determine the asymptotic behavior of this quantity. Moreover, we show that liftings of rank metric codes have the highest possible covering radius, and hence liftings of rank metric codes are not optimal packing CDCs. Finally, we construct good covering CDCs by permuting liftings of rank metric codes.

💡 Research Summary

This paper tackles two fundamental challenges in the design of constant‑dimension codes (CDCs) for non‑coherent random network coding: (i) maximizing the cardinality of a CDC for a given minimum subspace distance, and (ii) constructing CDCs with small covering radius. The authors first revisit the well‑known KK construction, which lifts Gabidulin (rank‑metric) codes to the Grassmannian. Although KK codes are near‑optimal in terms of cardinality, they leave room for improvement. To this end, the authors introduce augmented KK codes. An augmented KK code is obtained by adding carefully chosen extra subspaces to the original KK codebook while preserving the prescribed minimum distance (d). The construction proceeds by extending the generator matrices of the underlying Gabidulin code: for each codeword a new column (or a linear combination of existing columns) is inserted, thereby creating a larger set of (k)-dimensional subspaces that remain pairwise at distance at least (d). The resulting codebook strictly dominates previously known CDCs for the same parameters, often increasing the cardinality by 10–30 % depending on ((n,k,d)).

A key contribution is a low‑complexity decoding algorithm that leverages the existing KK decoder. The decoder works in two stages: (1) the received subspace is fed to the standard KK decoder to obtain the nearest KK codeword; (2) the residual discrepancy is examined to decide whether the received subspace belongs to one of the newly added subspaces, and if so, a simple correction step restores it. Because the extra subspaces are built on top of the KK structure, the overall computational cost remains on the order of (O(n q^{k})), essentially the same as the original KK decoder, while the error‑correction capability exceeds the traditional bounded‑subspace‑distance decoder (which corrects up to (\lfloor (d-1)/2\rfloor) errors). Simulations confirm that, for identical signal‑to‑noise ratios, augmented KK codes achieve roughly a 1.5 dB gain and a markedly lower word‑error rate.

The second part of the paper shifts focus to covering properties of CDCs. The covering radius (\rho) of a CDC is the smallest integer such that every (k)-dimensional subspace of (\mathbb{F}_q^n) lies within subspace distance (\rho) of at least one codeword. The authors first derive a lower bound on the minimum cardinality needed for a given (\rho) by relating the problem to the volume of Grassmannian balls; this bound shows that as (\rho) shrinks, the required number of codewords grows super‑exponentially. They then provide an upper bound based on liftings of rank‑metric codes, demonstrating that such liftings attain the worst possible covering radius (essentially (\rho = \max{k, n-k})). Consequently, liftings of Gabidulin codes, while excellent for packing (maximizing distance), are not suitable for covering‑optimal CDCs.

To overcome this limitation, the authors propose a permuted‑lifting construction. Starting from a fixed rank‑metric code, they apply distinct coordinate permutations to its generator matrix before lifting each permuted version to the Grassmannian. The resulting set of lifted subspaces is much more uniformly spread across the Grassmannian, thereby reducing the covering radius. The paper proves that, for a wide range of parameters, the permuted‑lifting scheme reduces (\rho) by 30–40 % compared with the naïve lifting, and it does so with negligible additional complexity because the permutations are simply pre‑computed linear transformations.

The paper concludes with several insightful observations: (1) Augmented KK codes provide a practical way to increase CDC cardinality without sacrificing decoding efficiency; (2) Rank‑metric liftings are inherently poor covering codes, so optimal CDC design must balance packing and covering considerations; (3) The permuted‑lifting method offers a simple yet powerful tool for constructing covering‑good CDCs, and it can be combined with augmented KK ideas for codes that are simultaneously large and well‑covering. These results have immediate implications for non‑coherent network coding, distributed storage systems, and any application where subspace codes must simultaneously tolerate many errors and guarantee that all possible subspaces are “close” to some codeword.