多约束最短链路分离路径精确算法

Exact Algorithm for Multi-Constrained Shortest Link-Disjoint Paths

在通信的源和目的间寻找两条(主用和备用)链路分离的QoS路径是提供可靠QoS路由的重要途径.现有求解多约束链路分离路径对(multi-constrained link-disjoint path pair,简称MCLPP)的算法难以保证求得存在于任意网络中的可行解和最优解.为解决这一问题,分析了MCLPP问题最优解的性质,提出了精确算法的设计原则,在此基础上给出了求解MCLPP问题的精确算法(link-disjoint optimal multi-constrained paths algorithm,简称LIDOMPA算法),可对任意网络求解客观存在的多约束最短链路分离路径对.为了降低算法的复杂性,引入了候选最优解、紧缩的约束向量和结构化的路径支配3种关键方法,在保障算法精确性的同时,有效地降低了LIDOMPA的搜索空间.大量的实验结果表明,LIDOMPA的求解能力优于现有算法,同时可以实现较低的算法执行时间开销.

Finding two link-disjoint QoS paths （primary and backup） between source-destination pairs is one of the most significant schemes to provide reliable QoS routing. Current algorithms for seeking multi-constrained link-disjoint path pair （MCLPP） can not always make sure to find the feasible solutions in networks. To solve this problem, this paper analyzes the properties of the optimal solution of MCLPP problem, and then proposes a design principle for the exact algorithm. Based on the design principle, an exact algorithm called link-disjoint optimal multi-constrained paths algorithm （LIDOMPA） is presented, which is able to find multi-constrained shortest link-disjoint path pair for arbitrary networks. To reduce the complexity, this paper introduces three key concepts： The candidate optimal solution, the constricted constraint vector and the structure-aware non-dominance, which effectively reduce the search space of LIDOMPA without loss of exactness. Extensive experiments show that LIDOMPA outperforms the existing algorithms in terms of the ability of obtaining solutions and achieves acceptable running time overhead.