Reinforcement learning negotiation strategy based on opponent classification

To help negotiation agent select its best actions and reach its final goal, this paper proposes a reinforcement learning negotiation strategy based on opponent classification. In the middle of negotiation process, negotiation agent makes the best use of the opponent´s negotiation history to make a decision of the opponent´s type, dynamically adjust the negotiation agent´s belief of opponent in time, and get more favorable and better negotiation result. Finally, the algorithm is proved to be effective and practical by experiment.