AdSCHE: DESIGN OF AN AUCTION-BASED FRAMEWORK FOR DECENTRALIZED SCHEDULING

Decentralized scheduling is one of the newly emerged avenues in scheduling research. It is concerned with allocating resources to alternative possible uses over time, where competing uses are represented by autonomous agents. Compared with classical scheduling models, decentralized scheduling is characterized with the distribution of scheduling knowledge and control, which introduces new levels of complexities, namely the coordination complexity due to the interaction problems among agents and the mechanism design complexity due to the self-interested nature of agents. These complexities intertwine and need to be addressed concurrently. This paper presents an auction-based framework which tackles coordination and mechanism design complexities through integrating an iterative bidding protocol, a requirement-based bidding language, and a constraint-based winner determination approach. Without imposing a time window discretization on resources the requirement-based bidding language allows bidders to bid for the processing of a set of jobs with constraints. Prices can be attached to quality attributes of schedules. The winner determination algorithm uses a depth-first branch and bound search. A constraint directed scheduling procedure is used at each node to verify the feasibility of the allocation. The bidding procedure is implemented by an ascending auction protocol. Experimental results show that the proposed auction framework exhibits improved computational properties compared with the general combinatorial auctions. A case study of applying the framework to decentralized media content scheduling in narrowcasting is also presented.