VeNAS: Versatile Negotiating Agent Strategy via Deep Reinforcement Learning (Student Abstract)

Takahashi, Toki; Higa, Ryota; Fujita, Katsuhide; Nakadai, Shinji

doi:10.1609/aaai.v36i11.21669

Cited by 5 publications

(6 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We assume a bilateral multi-issue negotiation, and employ the same negotiation environment as in the previous research (Takahashi et al 2022). In addition, to apply machine learning to negotiation agents, it is necessary to formulate Markov Decision Process (MDP) for multi-issue negotiations.…”

Section: End-to-end Scalable Negotiating Agent Strategy Via Multi-iss...mentioning

confidence: 99%

“…1 http://web.tuat.ac.jp/ ∼ katfuji/ANAC2021/ protocol (AOP) using finite MDP defined in the previous research (Takahashi et al 2022) Proposed Architecture Figure 1 illustrates the proposed reinforcement learning architecture via a multi-issue policy network. The environment includes the history of the bids exchanged between the agent and the opponent, including their own utility functions.…”

Section: End-to-end Scalable Negotiating Agent Strategy Via Multi-iss...mentioning

confidence: 99%

“…Automated negotiation strategies have attracted significant research attention, and competitions, such as Automated Negotiation Agents Competitions (ANAC) 1 , have been organized to discuss various negotiation strategies. The first comprehensive negotiation strategy is the versatile negotiating agent strategy (VeNAS) via deep reinforcement learning (Takahashi et al 2022). However, the limitation of VeNAS is that it has not demonstrated effective performance in the case of negotiation with a domain of large size.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Scalable Negotiating Agent Strategy via Multi-Issue Policy Network (Student Abstract)

Shimizu

Higa

Takahashi

et al. 2023

AAAI

View full text Add to dashboard Cite

Previous research on the comprehensive negotiation strategy using deep reinforcement learning (RL) has scalability issues of not performing effectively in the large-sized domains. We improve negotiation strategy via deep RL by considering an issue-based represented deep policy network to deal with multi-issue negotiation. The architecture of the proposed learning agent considers the characteristics of multi-issue negotiation domains and policy-based learning. We demonstrate that proposed method achieve equivalent or higher utility than existing negotiation agents in the large-sized domains.

show abstract

Section: End-to-end Scalable Negotiating Agent Strategy Via Multi-iss...mentioning

confidence: 99%

Section: End-to-end Scalable Negotiating Agent Strategy Via Multi-iss...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Scalable Negotiating Agent Strategy via Multi-Issue Policy Network (Student Abstract)

Shimizu

Higa

Takahashi

et al. 2023

AAAI

View full text Add to dashboard Cite

show abstract

“…Recently, agent negotiation strategies considering reinforcement learning have attracted attention because of their ability to adapt to various scenarios and opponents [27][28][29].…”

Section: Related Workmentioning

confidence: 99%

Multi-Issue Negotiation Protocol with Pre-Domain Narrowing

Hosokawa¹,

Kawata²,

Fujita³

2023

Applied Sciences

Self Cite

View full text Add to dashboard Cite

Consensus building among agents is crucial in multi-agent systems because each agent acts independently according to its utility function, and conflict among agents can occur. Therefore, automated negotiation is an essential technology for efficiently resolving conflicts and forming consensuses while also maintaining agents’ privacy. As the domain to be negotiated is large, the computational cost of reaching a consensus increases and the agreement rate decreases. Some negotiation protocols have been proposed wherein a mediator collects the utility information of each agent and creates multiple alternatives of agreements to handle large-scale multi-issue negotiations. However, in such protocols, a limitation is placed on agents’ privacy because all agents have to disclose their private information by following the mediator and pre-decided negotiation rules. In this study, we propose a negotiation protocol with a predomain-narrowing phase to enable efficient negotiations in large-scale domains which can maintain the privacy of information that agents should not disclose to their opponents or the mediator. The proposed protocol divides the negotiation process into a predomain-narrowing phase and the main negotiation phase. In the proposed protocol, the parts subject to negotiation are first narrowed in upon, and then the main negotiation is performed. We also propose two narrowing methods: issue- and option-narrowing. Further, we propose naive agent strategies considering the predomain-narrowing phase. We perform comparative simulation experiments between the baseline negotiation protocol without a domain-narrowing phase and the proposed negotiation protocol with the predomain-narrowing phase. The experimental results show that the proposed protocol achieves higher agreement rates in less negotiation time than the baseline.

show abstract

“…In recent automated negotiation research, some existing studies using reinforcement learning for agent bidding strategy have been investigated (Refs. [8], [9], [10], [11] etc.). On the other hand, most of the acceptance strategies in existing works are based on the heuristics approach [12] that do not use machine learning.…”

Section: Introductionmentioning

confidence: 99%

Effective Acceptance Strategy Using Deep Reinforcement Learning in Bilateral Multi-issue Negotiation

Matsuo,

Fujita

2024

Journal of Information Processing

Self Cite

View full text Add to dashboard Cite

Recently, automated negotiation has been attracting attention in multi-agent systems to resolve conflicts and reach an agreement among agents. In automated negotiation, two main types of strategies are incorporated in each agent: a bidding strategy that considers what kind of bid to send to an opponent, and an acceptance strategy that considers whether to accept the opponent's offer. In most bilateral multi-issue negotiation, agents take turns sending bids to each other and the negotiation ends when an agent accepts an opponent's offer. Therefore, the acceptance strategy is important in terms of increasing the utility of an agent. However, most studies of automated negotiation using reinforcement learning focus only on the bidding strategy of the agent, so there are not many studies that investigate acceptance strategies using reinforcement learning. In this paper, we propose a new configuration of a deep reinforcement learning framework for the acceptance strategy in automated negotiations using Deep Q-Network. The training phase is performed multiple times with various reward functions, and the reward capable of a higher utility value is investigated. Simulation experiments with other negotiating agents showed that the proposed method obtained significantly higher utility values than existing methods.

show abstract

VeNAS: Versatile Negotiating Agent Strategy via Deep Reinforcement Learning (Student Abstract)

Cited by 5 publications

References 1 publication

Scalable Negotiating Agent Strategy via Multi-Issue Policy Network (Student Abstract)

Scalable Negotiating Agent Strategy via Multi-Issue Policy Network (Student Abstract)

Multi-Issue Negotiation Protocol with Pre-Domain Narrowing

Effective Acceptance Strategy Using Deep Reinforcement Learning in Bilateral Multi-issue Negotiation

Contact Info

Product

Resources

About