Abstract
M.Sc. (Computer Science)
Artificial Intelligence is becoming increasingly relevant in everyday scenarios, from search engines to mobile phone assistants. Intelligent agents are an area of study in artificial intelligence where an agent is an entity that can act on its environment. Agents demonstrate qualities such as autonomy, situational awareness, embodiment, and flexibility.
A challenge faced is creating a mechanism for artificial intelligent agents to cross-communicate in a world where technology is always changing. This constant change leads to numerous problems, from changing protocols and deprecated communication endpoints to information or functionality loss.
The dissertation is an investigation into the existing frameworks leveraged in agent programming and the standards in place such as the Foundation for Intelligent Physical Agents standard, and the Mobile Agent System Interoperability Facility. These standards are detailed and expanded to include how to leverage modern technologies. The investigation provides a model or Decoupled Environment Agent Model that abstracts agent functioning from environment execution that is based on the existing standards. In this model, all components function as if the component is an agent. The model focuses on common interfaces and mechanisms for communication and discovery. The Decoupled Environment Agent Model is tested using an implementation of modern technologies, and aims to provide a mechanism where the various components of the model behave in a consistent simple manner.
The information gathered in the research, the model and the results collected while testing the prototype, are leveraged to identify whether an agent can effectively be loosely coupled from the environment. The implementation shows that it is possible to model an agent that is decoupled from the functioning environment, although there are many challenges to be faced with existing standards and the constantly changing nature of software today.