BDI代理人的意圖排程

Abstract

代理人系統是目前在計算機科學方面的最活躍的研究領域之一。 在各種代理人架構中，「信念、欲望、意圖」(BDI) 是最成熟的模型。 目前，關於 BDI 代理人的研究，大多數集中於代理人動態選擇計畫的能力， 即從欲望到意圖的能力。 而在這些被選擇的計畫應該被執行順序上，卻有很少有討論。 我們稱此問題為「意圖排程」。 在開發具有移動力和多重代理人的系統時，我們發現這些代理人必須能適應動態的變化。在缺乏適當的意圖排程時，代理人可能導致重複不必要的工作，浪費有價值的資源或者甚至導致使用者的計劃失敗。 因此，我們要研究一個有效的意圖排程方式，並且提出 BDI 推理過程裡一些有用的架構，使 BDI 代理人更有效率並且更適用於動態的環境。

Agent systems are currently one of the most active research fields in the computer science community. Among various agent architectures, BDI is probably the most mature model. Currently, most of the researches concerning BDI agents are focused on the ability for agents to dynamically select plans in order to achieve some goals, i.e. from Desire to Intention. However, there are few discussions on the order in which these selected plans should be executed. The problem might be called as Intention Scheduling. After developing systems with multiple agents of mobility and intelligence, we have discovered that these agents must adapt to dynamical and unpredictable changes. Without a proper scheduling, agents may result in repeating unnecessary work, wasting valuable resource or even failing the users’ expect altogether. Therefore, we are going to study an effective intention-scheduling scheme and some useful structures in BDI reasoning process, to make BDI agent more efficient and suitable for dynamic environment.