COMBIMA: Truthful, Budget Maintaining, Dynamic Combinatorial Market

Rica Gonen, Ozi Egri

Abstract: One-sided auctions have long been studied in economics and the computer science multi-agent planning domain [10, 12, 21]. Onesided auctions aim to find a high-social welfare (SWF) (efficient) allocation of a commodity to a set of agents, while ensuring that agents' best strategy is to truthfully report their input. An important extension of one-sided auctions are one-sided combinatorial auctions [17, 19, 20] where multiple commodities are offered for sale. Agents bid on bundles of commodities, which allows agents to express complex preferences over subsets of commodities (see [9] for many examples within). An elegant and well-studied class of combinatorial one-sided auctions are the sequential posted price auctions in which agents are presented sequentially with a vector of prices and must choose their preferred bundle given the price vector (among the first studied are [1, 18]). One-sided combinatorial auctions have been applied to various problems, including airport time-slot allocation [17], distributed query optimization [20] and transportation service procurement [19].