Sharper senses, sharper minds: Improving agent capabilities through better perception
Sebastian Welsh, "Sharper senses, sharper minds: Improving agent capabilities through better perception", Master of Science in Internetworking, UTS, Faculty of Information Technology, 2004.
Signcant enhancements in the capabilities of software agents can result through improving the way they acquire information. Decision making depends on getting the right information, but the issue of what actually constitutes the right information is complex. This research examines important characteristics of information that any decision maker should be able to acquire about their world, and uses this analysis to investigate how to improve the effectiveness of information acquisision in AI systems. As current agent implementations place limited emphasis on sensing, several mechanisms to improve sensing in agents are proposed.
A narrative approach is used to investigate what constitutes good information. This approach has been developed to speccally address this issue, as there is little previous investigation into mechanisms to identify factors that make particular information useful. The analysis identes that context and needs determine relevance, but within the relevant pool of information the decision maker needs information about objects, usage, places, participants and relationships.
An investigation of information acquisition in virtual worlds is conducted by examining design documentation for a range of virtual world implementations. These approaches are analysed to determine how well they supply good information and what constraints the approaches place on information acquisition. This analysis shows that existing approaches do not emphasise information acquisition and are subject to substantial constraints in how information is acquired. These constraints often result from design decisions about the decision architecture of the AI in combination with the problem domain of the setting, along with technical and storyline constraints.
The difference between the information that could be available and the information that is available in virtual worlds suggests a number of enhancements that can be made to enhance its usefulness and acquisition mechanisms. These include enriching the environment with information that matches the capabilities of the actors, and providing mechanisms tolter sensory input according to need, which ensures that the actor can identify the relevant aspects of its environment.
Developing a focus on information acquisition that is well articulated with decision making mechanisms provides substantial opportunities to improve the operation of software agents. The approach detailed in this research identes ways of enhancing the xibility and autonomy of agents without substantially adding to their complexity.
Interoperability in heterogeneous medical information systems using smart mobile agents and HL7
Bhavna Orgun, "Interoperability in heterogeneous medical information systems using smart mobile agents and HL7", MSc(Hons) Thesis. Macquarie University, Department of Computing, 2003.
Modern medical information infrastructure consists of many heterogeneous systems with diverse mechanisms to manage the underlying data. Patient information for a single individual might be spread across various systems due to several causes including patient relocation, and multiple treatments. There is a need for the ability to access consolidated patient data from different locations. This in turn necessitates a highly interoperable architecture.
...
A Design framework for implementing a configurable business application
Suharto Heng, A Design framework for implementing a configurable business application", Postgraduate Masters thesis, Macquarie University, 2001.
Most implementations of Business Application require major code rework when Business Requirement changes. This study addresses the possibility in using a design framework that may help developers to implement a Configurable Business Application. The goal of this design framework is to reduce the amount of code rework, and hence, reduced time to implement these changes.
Synthetic characters for simulated realistic environments world
Matthew Roberts, "Synthetic characters for simulated realistic environments world", BSc(hons) thesis, Macquarie University, Department of Computing, 2001.
We have developed a simulation environment called CreatureSpace that allows testing our agent theories on intelligent agents in a complex, dynamic realistic environment. The agents that populate this simulation environment have been crated using sketchy planning to deal with the uncertainty in the realistic environment and embedded knowledge to deal with complexity. We focus on how agents can manage large amounts of information and describe our embedded knowledge solution. We present the CreatureSpace architecture and our experiences in combining multiple artificial intelligence techniques in a uniform environment. The outcomes of the project are demonstrated through the running of fire evacuation simulations. In the process of creating these demonstrations we find that sketchy planning and embedded knowledge facilitate simple solutions to potentially difficult problems. Half- Life is used as the rendering engine for CreatureSpace, providing flexibility and strong graphics performance.
Contextualised learning through multimedia
Mita Bedi, "Contextualised learning through multimedia", BSc(hons) thesis, The University of New South Wales, School of Computer Science and Engineering, 2001.
There is great urgency about finding new and more effective ways through which learning can take place as both industry and society become dependent on educating effectively and efficiently. There are claims that the present education systems do not focus on the real goals of learning and using knowledge, rather there is a focus on artificial goals of assessments. Schools must concentrate on learning and teaching, not testing and comparing as this leads to diminishing interest to learn. The aim of this thesis is to establish a framework that enables the construction of these desirable learning environments. This framework will contain the attributes that are required to build such environments. An example of such an application "Interesting Money" will be built to demonstrate the framework and its effectiveness.
Contextualised learning through multimedia
Annie Y. S. Lau,"Contextualised learning through multimedia", BSc(hons) thesis, The University of New South Wales, School of Computer Science and Engineering, 2000.
Research shows that increasingly number of people are unable to transfer knowledge learnt from textbooks into everyday contexts. There is growing demand to learn and pressure to identify effective methods of learning is increasing. The aim of this thesis is to set up a framework for building self directed learning environments that enable effective learning. To demonstrate this, an educational environment on electronic auction called .e-Auction Investigator. has been developed. This framework is based on the educational paradigm, goal based-scenario (GBS), where the user is engaged in a scenario to achieve a goal. During this process, users participate in a scenario, perform tasks, make decisions and determine their path in the context. To assist with the decision making, users are encouraged to explore expert information that is available to them in multimedia format. It is expected that by completing the task, users would have been exposed to sufficient practical experience and working knowledge in the domain.
Peter Histon & Eugene Arena "Bus around Sydney", BSc(hons) thesis, The University of New South Wales, School of Computer Science and Engineering, 1999.
Making provision for the preferences of users is an important element of software and algorithm design. We make the claim that software that takes user preferences into account is more flexible, and therefore more usable.
A technique for handling user preferences is introduced. The technique is generic and can be applied to a variety of situations where user preferences are to be taken into account. This technique is demonstrated within the context of a route finding problem.
The route finding problem has been provided by Sydney Buses and involves getting between two places using buses that are a part of the Sydney Buses network. We present an algorithm that takes user preferences into account and that solves this problem. The algorithm presented is a variation of the A* search algorithm, extended to handle user preferences. Problems such as choosing which is the “best” route: a route with one interchange that takes five minutes or a route with no interchanges that takes ten minutes, are addressed and are solved using our technique for handling user preferences.
An interface that we have designed that enables users to specify their particular preferences is also presented. While this interface has been developed specifically for the Sydney Buses problem, it is universal enough such that the main ideas behind it can be reused for other preference-related problems.
John B. Zvonar,"Survival in a virtual world", BSc(hons) thesis, The University of New South Wales, School of Computer Science and Engineering, 1999.
In the ever-changing world of today’s online gaming worlds and chat channels, there is an increasing tendency to construct more interactive and immersive surroundings for users to participate in. To this extent, agents play a vital part in existing in this environment. As users, characters, and servants within their systems, these agents can interact with others in order to help, instruct or participate themselves. In this type of environment, typically an imitation of the real world itself, an agent not only has to exhibit an intelligent behavior, but also has to mimic that of a normal human’s.
We explore how to construct such an agent that can handle a complex, dynamic world of this nature, where humans themselves struggle to survive in a virtual existence. In essence, we wish to look for the simplest, modular solution possible for creating what appears to be an intelligent agent that can survive in this virtual world. This modular approach stands to give the agent intelligence in the combination of smaller aspects can be added to the agents behavior as a whole, whilst the simplicity of these aspects ensures that the interaction between modules is easily handled and modifiable. The culmination of this work is an agent that can exist in a artificial world such that it can not only survive, but also helps others to, and at the same time exhibit an intelligent behavior that is fitting for the role that it plays.