Optimal Parametrization and Comparative Analysis of Preallocation Methods for Combinatorial Auction-Based Channel Assignment

Algorithms based on combinatorial auction (CA) show significant potential regarding their application for channel assignment problems in multiconnective ultra-reliable wireless networks. However the computational effort required by such algorithms grows fast with the number of users and resources. Therefore, preallocation-based CA represents a promising approach for these setups. The aim of preallocation is to constrain the number of bids submitted by participants in the CA process, thus allow the numerical feasibility of the auction problem. Reduction of bid number is achieved via limiting the number of items (in our case channels) considered by auction participants (tenants) in their bids. Thus the aim of preallocation is to non-exclusively assign channels to tenants, which assignment serves as a basis for the later bid generation in the CA procedure. In this paper we compare the performance of various preallocation approaches via simulation according to various measures, namely the total utility of the resulting allocation, the number of unassigned channels and the required computational time. In addition to simple iterative random and semi-random algorithms which serve as baseline reference, we consider the many-to-many version of the Gale-Shapley algorithm (M2MGS) and introduce a relaxed version of the combinatorial auction algorithm (RCA) for the preallocation stage, which allows the multiple allocation of single items. Furthermore, we analyze the optimal parametrization of the M2MGS and the RCA preallocation methods, and formulate recommendations for optimal performance based on the analysis.