Time-Quality Tradeo s in Reallocative Negotiation with Combinatorial Contract Types

The capability to reallocate items|e.g. tasks, securities, bandwidth slices, Mega Watt hours of electricity, and collectibles|is a key feature in automated negotiation. Especially when agents have preferences over combinations of items, this is highly nontrivial. Marginal cost based reallocation leads to an anytime algorithm where every agent's payo increases monotonically over time. Di erent contract types head toward di erent locally optimal allocations of items, and OCSM-contracts head toward the global optimum. Reaching it can take impractically long, so it is important to trade o solution quality against negotiation time. To construct negotiation protocols that lead to good allocations quickly, we evaluated original (O), cluster (C), swap (S), and multiagent (M) contracts experimentally. O-contracts led to the highest social welfare when the ratio of agents to tasks was large, and C-contract were best when that ratio was small. O-contracts led to the largest number of contracts made. M-contracts were slower per contract, and required a signi cantly larger number of contracts to be tried to verify that a local optimum had been reached. S-contracts were not competitive because they restrict the search space by keeping the number of items per agent invariant. O-contracts spread the items across agents while C-contracts and M-contracts concentrated them on a few agents.