An efficient and robust negotiating strategy in bilateral negotiations over multiple items

https://doi.org/10.1016/j.engappai.2014.05.008Get rights and content

Abstract

Multi-item negotiations surround our daily life and usually involve two parties that share common or conflicting interests. Effective automated negotiation techniques should enable the agents to adaptively adjust their behaviors depending on the characteristics of their negotiating partners and negotiation scenarios. This is complicated by the fact that the negotiation agents are usually unwilling to reveal their information (strategies and preferences) to avoid being exploited during negotiation. In this paper, we propose an adaptive negotiation strategy, called ABiNeS, which can make effective negotiations against different types of negotiating partners. The ABiNeS strategy employs the non-exploitation point to adaptively adjust the appropriate time to stop exploiting the negotiating partner and also predicts the optimal offer for the negotiating partner based on the reinforcement-learning based approach. Simulation results show that the ABiNeS strategy can perform more efficient exploitations against different types of negotiating partners, and thus achieve higher overall payoffs compared with the state-of-the-art strategies under negotiation tournaments. We also provide a detailed analysis of why the ABiNeS strategy can negotiate more efficiently compared with other existing state-of-the-art negotiation strategies focusing on two major components. Lastly, we propose adopting the single-agent best deviation principle to analyze the robustness of different negotiation strategies based on model checking techniques. Through our analysis, the ABiNeS strategy is shown to be very robust against other state-of-the-art strategies under different negotiation contexts.

Introduction

Negotiation is a common and important approach to resolve conflicts and reach agreements between different parties in our daily life. With the advance and popularity of Web and e-commerce, it is expected that a lot of previous negotiation activities between humans will be moved to electric platforms and greatly benefit from automated negotiation techniques (Kersten and Noronha, 1999). Automated negotiation techniques can, to a large extent, alleviate the efforts of human negotiators, and also aid human in reaching better negotiation outcomes by compensating for the limited computational abilities of humans when they are faced with complex negotiations. Until now, a lot of automated negotiation strategies and mechanisms have been proposed in different negotiation scenarios (Faratin et al., 2003, Saha et al., 2005, Hindriks and Tykhonov, 2008, Jakub and Ryszard, 2006).

The major difficulty in designing automated negotiation agent is how to achieve optimal negotiation results given incomplete information on the negotiating partner. The negotiation partner usually keeps its negotiation strategy and its preference as its private information to avoid exploitations. A lot of research effort has been devoted to better understand the negotiation partner by either estimating the negotiation partner׳s preference profile (Zeng and Sycara, 1998, Hindriks and Tykhonov, 2008, Coehoorn and Jennings, 2004) or predicting its decision function (Zeng and Sycara, 1996, Jakub and Ryszard, 2006). On one hand, with the aid of different preference profile modeling techniques, the negotiating agents can get a better understanding of their negotiating partners and thus increase their chances of reaching mutually beneficial negotiation outcomes. On the other hand, effective strategy prediction techniques enable the negotiating agents to maximally exploit their negotiating partners and thus receive as much benefit as possible from negotiation. However, in most of previous work, the negotiating agents are usually assumed to be situated in a negotiation environment that strong limitations are put on the negotiation scenario or the negotiation opponent. For example, some work assumes (Saha et al., 2005) that the agents negotiate over a single item only, however practical negotiation scenarios usually involve multiple items to negotiate over. There also exists some work that assumes that the negotiation agents can have access to their opponents׳ (partial) preferences. This can be unrealistic especially in multi-issue negotiation scenarios in which the preferences of different agents may vary significantly, and the agents usually would not like to disclose their preferences to avoid being exploited. Another assumption commonly adopted (Jakub and Ryszard, 2006) is that the negotiation opponent is limited to choose from a specific set of simple strategies, e.g., time-dependent or behavior-dependent tactics. Those strategies designed under this assumption may not work well against other negotiation partners with more complex state-of-the-art strategies.

To this end, in recent years a number of advanced negotiation strategies taking advantage of existing techniques have been proposed and agents employing these strategies have participated in automated negotiating agents competition (ANAC) (Baarslag et al., 2010, Baarslag et al., 2011a). The ANAC competition provides a negotiation platform which enables different negotiation agents to be evaluated against a wide range of opponents within a realistic negotiation environment. During the past three years, dozens of state-of-the-art negotiation strategies have been extensively evaluated in a variety of multi-issue negotiation scenarios and valuable insights have been obtained in terms of the advantages and disadvantages of different techniques, e.g., the efficacy of different acceptance conditions (Baarslag et al., 2011). It is still an open and interesting problem to design more efficient automated negotiation strategies against a variety of negotiating opponents in different negotiation domains.

In this paper, we propose an adaptive negotiation strategy ABiNeS for automated agents to negotiate in bilateral multi-issue negotiation environments following the settings adopted in ANAC 2012 (The Third International Automated Negotiating Agent Competition, 2012). Bilateral multi-issue negotiations surround people׳s daily life and have received lots of attention in the negotiation literature. During negotiation, both the agents׳ negotiation strategies and preference profiles are their private information, and for each agent the only available information about the negotiating partner are its past negotiation moves. Considering the diversity of the available negotiation strategies that the negotiating agents can adopt, it is usually very difficult (or impossible) to predict which specific strategy the negotiating partner is using based on this limited information. To effectively cope with different types of opponents, we introduce the concept of non-exploitation point λ to adaptively adjust the degree that an ABiNeS agent exploits its negotiating opponent. The value of λ is determined by the characteristics of the negotiation scenario and the concessive degree of the negotiating partner, which is estimated based on the negotiation history. Besides, to maximize the possibility that the offer the ABiNeS agent proposes will be accepted by its negotiating partner, it can be useful to make predictions on the preference profile of the negotiating partner. Instead of explicitly modeling the negotiation partner׳s preference profile, we propose a reinforcement-learning based approach to determine the optimal proposal for the negotiating partner based on the current negotiation history.

We evaluate the performance of the ABiNeS strategy compared with a number of state-of-the-art negotiation strategies from two different perspectives: efficiency in terms of the average payoff obtained under a particular negotiation tournament and robustness in terms of how likely the agents have the incentive to adopt our strategy rather than other strategies. First, the efficiency evaluation is conducted under the negotiation tournament setting following ANAC 2012 using GENIUS1 (Lin et al., 2012) platform. Simulation results show that the ABiNes strategy can make more effective exploitations against a variety of negotiation partners and thus obtain higher average payoffs during negotiation tournaments and it is worth mentioning that the ABiNes strategy wins the champion of ANAC 2012 known as CUHKAgent. Second, we give a detailed analysis of the ABiNes strategy by investigating the influence of its two major novel components. Through the detailed analysis, we aim at providing a clear understanding of why the ABiNes strategy can win the champion of ANAC 2012, and more importantly, offering valuable insights for the automated negotiation community for the future negotiation strategy design. Third, we propose adopting the single-agent best deviation principle to analyze the robustness of different negotiation strategies based on model checking techniques. According to the robustness analysis, the ABiNes strategy is shown to be very robust against other state-of-the-art strategies under different negotiation contexts.

The remainder of the paper is organized as follows. In Section 2, we give a description of negotiation model we consider in this paper. In Section 3, the negotiation strategy ABiNes we propose is introduced. In Section 4, we give detailed evaluation of the negotiation efficiency and robustness of ABiNes compared with the state-of-the-art negotiation strategies under different negotiation contexts. An overview of related work on automated negotiation strategies is given in Section 5. Lastly conclusion and future work are given in Section 6.

Section snippets

Negotiation model

In this section, we describe the negotiation model we consider in this work, which follows the settings adopted in ANAC 2012 (The Third International Automated Negotiating Agent Competition, 2012). We focus on bilateral negotiations, i.e., negotiations between two agents. Specifically, the alternating-offers protocol is adopted to regulate the interactions between the negotiating agents, in which the agents take turns to exchange proposals. For each negotiation scenario, both agents can

ABiNeS: an adaptive bilateral negotiating strategy

In this section, we describe the adaptive negotiating strategy ABiNeS in detail. For the ease of description, we refer to the ABiNeS agent as agent A and its negotiating partner as agent B in the following descriptions. The ABiNeS strategy mainly consists of four basic decision components. The first component is Acceptance-Threshold (AT) component and it is responsible for determining the ABiNeS agent׳s minimum acceptance threshold lAt at time t. The second component is Next-Bid (NB) component

Experimental evaluation and analysis

In this section, we evaluate the negotiation performance of the ABiNeS strategy against the state-of-the-art negotiation agents across a wide range of multi-issue negotiation scenarios in the tournament setting following previous evaluation criteria (i.e., efficiency and robustness) (Baarslag et al., 2010, The Second International Automated Negotiating Agent Competition, 2011, The Third International Automated Negotiating Agent Competition, 2012) under the GENIUS (General Environment for

Related work

Until now great effects have been devoted to develop efficient negotiation strategies for automated negotiating agents in the literature. The most typical approach in the literature is concession-based strategy such as ABMP strategy (Jonker et al., 2006). The ABMP strategy agent decides on the next move based on its own utility space only and makes concession to its negotiating partners according to certain concession pattern. However, since the dynamics and influences of the negotiating

Conclusion and future work

In this paper, we propose an adaptive negotiation strategy ABiNeS for automated agents to negotiate in bilateral multi-issue negotiation scenarios. We introduce the concept of non-exploitation point λ to adaptively adjust the ABiNeS agent׳s concession degree to its negotiating opponent, and propose a reinforcement-learning based approach to determine the optimal proposal for the negotiating partner to maximize the possibility that the offer will be accepted by the opponent.

The performance of

Acknowledgments

This work is supported by National Natural Science Foundation of China (NSFC) 61170077, SZ-HK Innovation Circle project ZYB200907060012A, NSF GD: 10351806001000000, GDS&T 2012B091100198, S&T project of SZ JC200903120046A, and CUHK Direct Grant for Research (Project ID 4055024).

References (30)

  • G.E. Kersten et al.

    Www-based negotiation supportdesign, implementation, and use

    Decis. Support Syst.

    (1999)
  • D. Zeng et al.

    Bayesian learning in negotiation

    Int. J. Human Comput. Syst.

    (1998)
  • T. Baarslag et al.

    The first automated negotiating agents competition (anac 2010)

    New Trends in Agent-Based Complex Automated Negotiations

    (2010)
  • Baarslag T, Fujita K, Gerding E H, et al. Evaluating practical negotiating agents: Results and analysis of the 2011...
  • Baarslag, T., Hindriks, K., Jonker, C., 2011. Acceptance conditions in automated negotiation. In:...
  • Chen, S.Q., Weiss, G., 2012. An efficient and adaptive approach to negotiation in complex environments. In: Proceedings...
  • E.M. Clarke et al.

    Model checking and abstraction

    ACM Trans. Program. Lang. Syst.

    (1994)
  • Coehoorn, R.M., Jennings, N.R., 2004. Learning an opponent׳s preferences to make effective multi-issue negotiation...
  • P. Faratin et al.

    Using similarity criteria to make negotiation trade-offs

    Artif. Intell.

    (2003)
  • Fatima, S.S., Wooldridge, M., Jennings, N.R., 2007. Approximate and online multi-issue negotiation. In: Proceedings of...
  • K. Fujita et al.

    An approach to scalable multi-issue negotiationdecomposing the contract space

    Comput. Intell.

    (2012)
  • Hao, J.Y., Song, S.Z. Online Technical Report....
  • Hindriks, K., Tykhonov, D., 2008. Opponent modeling in auomated multi-issue negotiation using Bayesian learning. In:...
  • Hindriks, K., Jonker, C.M., Tykhonov, D., 2006. Eliminating interdependencies between issues for multi-issue...
  • Jakub, B., Ryszard, K., 2006. Predicting partner׳s behaviour in agent negotiation. In: AAMAS ׳06, pp....
  • Cited by (0)

    View full text