Evolving Best-Response Strategies for Market-Driven Agents Using Aggregative Fitness GA

There are very few existing works that adopt genetic algorithms (GAs) for evolving the most successful strategies for different negotiation situations. Furthermore, these works did not explicitly model the influence of market dynamics. The contribution of this work is developing bargaining agents that can both: 1) react to different market situations by adjusting their amounts of concessions and 2) evolve their best-response strategies for different market situations and constraints using an aggregative fitness GA (AFGA). While many existing negotiation agents only optimize utilities, the AFGA in this work is used to evolve best-response strategies of negotiation agents that optimize their utilities, success rates, and negotiation speed in different market situations. Given different constraints and preferences of agents in optimizing utilities, success rates, and negotiation speed, different best-response strategies can be evolved using the AFGA. A testbed consisting of both: 1) market-driven agents (MDAs)—negotiation agents that make adjustable amounts of concessions taking into account market rivalry, outside options, and time preferences and 2) GA-MDAs—MDAs augmented with an AFGA, was implemented. Empirical results show that GA-MDAs achieved higher utilities, higher success rates, and faster negotiation speed than MDAs in a wide variety of market situations.