在智慧家庭中,系統為了滿足不同使用者的需求可能會面臨意見不合或資源供不應求等衝突狀況,智慧家庭會根據使用者的偏好來提供個人化的服務,不同的偏好可能會對彼此產生干擾或是爭奪系統資源的問題,因此必須透過協商的方法尋求彼此的共識。自動協商系統在這樣的前提下扮演著關鍵的角色,藉由搜尋滿足使用者的解答,以解決衝突或資源限制的問題;此外,進行協商時,參與者在考量利益權衡的狀況下,會將自身的個人資訊隱藏起來,在這樣的條件之下便會增加協商的難度。本研究提出一種適用於非合作式的多重代理人、多議題的協商模型,在協商過程中使用連續投影策略的方法,即使缺乏對手的完整資訊依然能夠達成協議。在智慧家庭的環境下,透過BOA架構使代理人皆成為協商代理人並且依據其使用者的個人偏好在協商過程中爭取利益。本研究開發一個系統讓多重代理人在連續提案的機制下進行協商,實現連續投影策略的方法,在不完整資訊的環境下也能找出滿足各使用者的接受條件,進而達成協議,使結果能符合使用者的需求。
For satisfying the needs of different users in a smart home, the issue of conflict resolution must be considered. Conflicts may come from disagreement in opinions and may also be caused by exceeding the usage of shared resources. A smart home provides personalized service based on preferences of users. Explicit or implicit preferences of different users may interfere with each other or compete for system resources. A negotiation method can be considered as designing a process to search a consensus among each other. For searching a solution which satisfies all users, automatic negotiation systems play a key role to resolve problems such as conflicts or inadequate resource assignments. Sometimes, participants in the negotiation process hide private information because of considering their benefits, and it makes negotiation more difficult. This thesis proposes an automatic multiple issue negotiation model suitable in non-cooperative multiagent. By using sequential projection strategies, the negotiation process always reaches an agreement, even lacking some information about opponents. Through the BOA architecture, every agent can become a negotiating agent which works for the benefits of the user who it represents. A prototype system has been developed in realizing the strategies in a multiagent setting.