An extended Belief-Desire-Intention Model for Human - Agent Collaboration
The Belief-Desire-Intention (BDI) model of agency as traditionally realised in software frameworks does not address human/agent collaboration. Rather, its focus is limited to autonomous goal directed behaviour for situated intelligent agents. Nonetheless, intelligent agents developed using BDI frameworks have been deployed in many challenging applications, such as space shuttle fault diagnosis, UAV control, air traffic control, manufacturing automation and large-scale military simulations (Jarvis, D et al. 2013). These deployments have been facilitated by the availability of commercial strength software frameworks, namely PRS (Procedural Reasoning System) (Georgeff & Lansky 1986), dMARS (distributed MultiAgent Reasoning System) (D'inverno et al. 2005), JACK Intelligent Agents (Agent Oriented Software Pty. Ltd. 2012b) and most recently GORITE (Jarvis, D et al. 2013; Rönnquist 2007). Indeed, BDI frameworks have become, along with Soar (Laird, Newell & Rosenbloom 1987) the framework of choice for the development of knowledge intensive agent applications (Jones & Wray 2006).
However, BDI frameworks have experienced limited success in their application to situations that require agents to interact with humans in the achievement of goals. A key reason for this has been that while traditional BDI agents use goals internally, the representation is not explicit and consequently it is difficult for humans to directly influence agent activity and vice-versa in a collaborative manner. With the advent of GORITE, this representational constraint has been removed and the possibility now exists for BDI agents to work collaboratively with humans in goal achievement and to facilitate the development of a new generation of intelligent decision support systems. In this thesis, we present a conceptual model for collaborative BDI agent behaviour that extends the BDI model of agency. The model is then validated by providing the GORITE framework with additional collaborative functionality and then developing a number of proof-of-concept applications using the extended framework.
History
Number of Pages
111Location
Central Queensland UniversityPublisher
Central Queensland UniversityPlace of Publication
Rockhampton, QueenslandOpen Access
- Yes
Era Eligible
- No
Supervisor
Dr. Jacqueline Jarvis and Dr. Dennis JarvisThesis Type
- Master's by Research Thesis
Thesis Format
- Traditional