Could not connect: Connection refused
• Swarm Intelligence: Crowd Dynamics • Swarm Intelligence: Aggregate Movement • Swarm Intelligence: Ant Systems • Social Systems: Trust and Reputation • Social Systems: Referral Networks • Swarm Intelligence: Control •
Swarm Intelligence: Crowd Dynamics
This project investigates the dynamics of crowds in confined spaces. The techniques used include hybrid agent-based modeling. Agents are simple, sensing multiple information fields within an environment. This work extends models by Kirchner and Helbing by introducing models for injury. The goals of this project are to develop agent communication techniques based upon local exchange of information that result in safer movement of crowds in confined spaces.
Swarm Intelligence: Aggregate Movement
This project attempts to answer the question, "Why is aggregate movement so stable?". Agent-based models are developed that replicate the qualitative characteristics of real fish. Models of zebra danio have been developed and compared to simple attraction-repulsion models to demonstrate that single sensor, point models cannot accurately replicate the dynamics of actual fish. The goals of this project are deep understanding of these complex systems and the ways in which they can be controlled along with the development of potential search and robotic applications.
Swarm Intelligence: Ant Systems
This project proposes and investigates improvements to ant-based systems. Several extensions to Ant Colony Optimization (ACO) algorithms have been proposed, most notably Local Based Tour (LBT). Applications and theory of ant systems are investigated. Task allocation algorithms in insect systems and their application to real time control are investigated along with the recruitment problem. The goals of this project are the development of domain specific search algorithms with potential application in swarm robotics.
Social Systems: Trust and Reputation
This project investigates attacks and mechanisms for defeating them in social systems where agents interact requesting and providing services. Agents use trust and reputation to decide when, how and with whom to interact. Agent-based modeling, game theory and other techniques are used to evaluate trust and reputation algorithms and develop new threat models. Con-resistant trust has been the most noteworthy development of this project leading to the identification of multi-dimensional trust as a desirable characteristic within trust and reputation models. The goals of this project are the development of trust and reputation frameworks that could be deployed in e-commerce and other domains.
Social Systems: Referral Networks
This project investigates the dynamics of systems in which agents interact with only their neighbors to request and provide services. Agents receiving requests can either provide a service or refer the requesting agent to another agent by providing a referral. This is the essence of a referral network. Extensions to referral networks in which service ontologies, or concept relations, are provided are investigated. It has been demonstrated that meaning is an emergent property of referral networks in which concept relations are shared. The goals of this project are the development of decentralized intelligent information systems that rely less on keyword search.
Swarm Intelligence: Control
This project involves the use of swarm intelligence techniques for control in networks. Examples of problems addressed are traffic control and influence control in information retrieval systems.
• Evolutionary Computation: Immune Systems • Evolutionary Computation: Genetic Algorithms/Programming • Software Paradigms: Mobile Code and Hot Swapping •
Evolutionary Computation: Immune Systems
Immune systems efficiently detect self and non-self components in a system using a set of dynamically changing detectors. This project has used Artificial Immune Systems (AIS) to demonstrate their utility in e-mail spam detection and time series analysis. In particular, the spam detection algorithms developed were competitive with Spam Assassin while using a significantly smaller classifier set. The goals of this project were the development of a framework for AIS for experimental evaluation of AIS algorithms.
Evolutionary Computation: Genetic Algorithms/Programming
Genetic Algorithms (GA) and Genetic Programming (GP) are algorithms modelled on the principles of genetics. These search techniques have been shown to be highly effective in searching large spaces where little or no domain knowledge is included in the search process. In this project we look at the performance of new operators, hybrid algorithms and parameter settings to improve performance. Applications are also investigated, mainly in the area of telecommunications. Some of the algorithms developed have resulted in patents.
Software Paradigms: Mobile Code and Hot Swapping
This project explores the use of mobile code in extending the behavior of distributed applications in real time. Investigations include exploration of algorithms for collaboration and coordination of mobile code along with the interfaces required for their management and control. As a subordinate activity, the problems of hot swapping parts of an application have been explored and how the issues presented by constantly evolving code can be applied to legacy code.