دانلود رایگان مقاله الگوریتم آنلاین مبتنی بر باند برای ساخت کدهای متعامد نوری

abstract

An optical orthogonal code (OOC) is a family of binary sequences with good auto- and cross-correlation properties. In the literature, various mathematical tools have been used to construct OOCs with specific parameters. But, to find a complete solution for constructing OOCs with an arbitrary setting of parameters is still difficult at the moment. In this paper, a clique-based online algorithm is proposed to construct OOCs of relatively large sizes. In the proposed algorithm, the construction of OOCs is reduced to the maximum clique problem based on specially generated graphs, where vertices represent the codewords of an OOC and edges represent the cross-correlation relationships between codeword pairs. In order to overcome the limitation of computer memory for storing large graphs, part of the graph vertices are supposed to arrive sequentially to be fed into the proposed algorithm, and a specially designed evolutionary algorithm is used to find the maximum clique of the current graph when new vertices arrive. The proposed algorithm does not use parameter-specific techniques and hence can be used for different code weight and correlation constraints. Experiments show that the proposed algorithm outperforms an offline evolutionary algorithm with guided mutation on constructing OOCs.

5. Conclusions and future work

Due to its great importance in the field of communications, the problem of designing OOCs has attracted increasing attentions in recent years. Many good results have been obtained by an extensive use of various mathematical tools. But, it seems that we cannot expect the problem to be completely solved for arbitrary parameters in the near future. Meanwhile, the design of OOCs can be reduced to the maximum clique problem (MCP) in graph theory, for which there are many successful heuristics. As a result, we propose clique-based heuristics to design OOCs with relatively large sizes. In this paper, we present an online algorithm for constructing OOCs based on an EA with guided mutation (EA/G) [33]. Note that the cardinality of the OOC graph is usually huge and the corresponding clique size is quite limited. Hence, the usual clique-based strategy may fail in this case. So, we suppose all the graph vertices arrive in an online manner, and present a substitution technique which leads to our EA with substitution (EAS). Experiments show that EAS has a similarly good performance to EA/G for the MCP on DIMACS benchmarks. Then, we use EAS to construct OOCs with different code weights and different correlation constraints. The experimental study shows that EAS returns OOCs of sizes close to the current best upper bound when the code length is small. Taking (v, 4, 2) for instance, the ratio of the obtained code size to the Johnson bound is over 90% when v < 100. If more running time is admitted, the result could be improved.