Previous attempts at implementing bots for the classic roguelike game NetHack have been hindered by many problems related to its complexity and console-based interface. The framework implemented as part of this work solves the problem of interfacing with the game and provides a programmer-friendly API for the Java and Clojure programming languages. It enables programming sophisticated bots using the provided model of the game world, a library of possible actions and utilities for various aspects of the game. The framework uses elements of functional and logic programming and doesn't require modifications of the game. Also described is an implementation of the first NetHack bot capable of winning the game. Powered by TCPDF (www.tcpdf.org)

Degree: Master of Science in Engineering Department: Engineering / Multi-Agent Systems are becoming a popular paradigm for many engineering applications. However, there is still much research to be performed in this fast growing field. In this thesis, the effect of learning in multi-agent systems on communication and collaboration between agents is investigated. This research focuses on agents learning local cooperative behaviour from a centralised agent, as well as using learning to reduce the amount of communication between agents that use negotiation to achieve their goals. A simple test problem is formulated in MATLAB. The effect of learning is clearly seen to reduce the amount of communication between agents by up to 50%, while still maintaining cooperative behaviour. The agents are also demonstrated to learn to a large degree cooperative local behaviour from a central system.

acase@tulane.edu / Chlamydia trachomatis (Ct) is the most common bacterial sexually transmitted infection (STI) in the United States and is major cause of infertility, pelvic inflammatory disease, and ectopic pregnancy among women. Despite decades of screening women for Ct, rates continue to increase among them in high prevalent areas such as New Orleans. A pilot study in New Orleans found approximately 11% of 14-24 year old of African Americans (AAs) were infected with Ct. Our goal is to mathematically model the impact of different interventions for AA men resident in New Orleans on the general rate of Ct among women resident at the same region. We create and analyze mathematical models such as multi-risk and continuous-risk compartmental models and agent-based network model to first help understand the spread of Ct and second evaluate and estimate behavioral and biomedical interventions including condom-use, screening, partner notification, social friend notification, and rescreening. Our compartmental models predict the Ct prevalence is a function of the number of partners for a person, and quantify how this distribution changes as a function of condom-use. We also observe that although increased Ct screening and rescreening, and treating partners of infected people will reduce the prevalence, these mitigations alone are not sufficient to control the epidemic. A combination of both sexual partner and social friend notification is needed to mitigate Ct. / 1 / Asma Aziz Boroojeni

本研究計畫的主題在於研究代理人的可信度管理，首要的目標是建立一個以代理人為基礎的安全式電子交易環境。以目前的情況來看，唯有代理人的觀念及技術來執行電子商務仲介者的角色，利用軟體代理者具有自主性，及適時反應等特質，提供服務時的效益和彈性，再輔以適當的安全性管理及深入的可信度探討，電子商務才可能被具體應用到人類實際日常生活上。在作法上除了採用FIPA的規格作為代理人系統平台的實作標準，延伸XML/RDF來便利代理人的建構與溝通，更進一步結合X.509及SPKI/SDSI兩種類型憑證的優點，導入分散式認證授權的觀念，並透過RBAC的控管，形成多重代理人系統的安全架構。配合相關的信任策略及商務模型，以期完成建構一個以代理人為基礎可信任安全式電子交易環境的目標。 / This thesis describes an agent-based secure E-Commerce environment with distributed authentication and authorization services. The previous researches about security issues in agent-mediated E-commerce do not solve the problems of deals with strangers. We merge role based access control (RBAC) concept for adapting the certificates to different business models or new content-based network. Several types of agent delegation mechanism based on our role certificates and some considerations about how to achieve agent trust management with policies both in logics and practice are presented. Finally, We will demonstrate a scenario on FIPA OS system by using agent communication language (ACL) and content language (CL) encoded by XML and XML/RDF.

One of the limitations of the BDI (Belief-Desire-Intention) model is the lack of any explicit mechanisms within the architecture to be able to learn. In particular, BDI agents do not possess the ability to adapt based on past experience. This is important in dynamic environments as they can change, causing previously successful methods for achieving goals to become inefficient or ineffective. We present a model in which learning, analogous reasoning, data pruning and learner accuracy evaluation can be utilised by a BDI agent and verify this model experimentally using Inductive and Statistical learning. Intelligent Agents are a new way of developing software applications. They are an amalgam of Artificial Intelligence (AI) and Software Engineering concepts that are highly suited to domains that are inherently complex and dynamic. Agents are software entities that are autonomous, reactive, proactive, situated and social. They are autonomous in that they are able to make decisions on their own volition. They are situated in some environment and are reactive to this environment yet are also capable of proactive behaviour where they actively pursue goals. They are capable of social behaviour where communication can occur between agents. BDI (Belief Desire Intention) agents are one popular type of agent that support complex behaviour in dynamic environments. Agent adaptation can be viewed as the process of changing the way in which an agent achieves its goals. We distinguish between 'reactive' or short-term adaptation, 'long-term' or historical adaptation and 'very long term' or evolutionary adaptation. Short-term adaptation, an ability that current BDI agents already possess, involves reacting to changes in the environment and choosing alternative plans of action which may involve choosing new plans if the current plan fails. 'Long-term' or historical adaptation entails the use of past cases during the reasoning process which enables agents to avoid repeating past mistak es. 'Evolutionary adaptation' could involve the use of genetic programming or similar techniques to mutate plans to lead to altered behaviour. Our work aims to improve BDI agents by introducing a framework that allows BDI agents to alter their behaviour based on past experience, i.e. to learn.

This thesis presents SodaBot, a general-purpose software agent user-environment and construction system. Its primary component is the basic software agent --- a computational framework for building agents which is essentially an agent operating system. We also present a new language for programming the basic software agent whose primitives are designed around human-level descriptions of agent activity. Via this programming language, users can easily implement a wide-range of typical software agent applications, e.g. personal on-line assistants and meeting scheduling agents. The SodaBot system has been implemented and tested, and its description comprises the bulk of this thesis.

The increasing number of network security related incidents has made it necessary for the organizations to actively protect their sensitive data with network intrusion detection systems (IDSs). Detecting intrusion in a distributed network from outside network segment as well as from inside is a difficult problem. IDSs are expected to analyze a large volume of data while not placing a significant added load on the monitoring systems and networks. This requires good data mining strategies which take less time and give accurate results. In this study, a novel hybrid layered multiagent-based intrusion detection system is created, particularly with the support of a multi-class supervised classification technique. In agent-based IDS, there is no central control and therefore no central point of failure. Agents can detect and take predefined actions against malicious activities, which can be detected with the help of data mining techniques. The proposed IDS shows superior performance compared to central sniffing IDS techniques, and saves network resources compared to other distributed IDSs with mobile agents that activate too many sniffers causing bottlenecks in the network. This is one of the major motivations to use a distributed model based on a multiagent platform along with a supervised classification technique. Applying multiagent technology to the management of network security is a challenging task since it requires the management on different time instances and has many interactions. To facilitate information exchange between different agents in the proposed hybrid layered multiagent architecture, a low cost and low response time agent communication protocol is developed to tackle the issues typically associated with a distributed multiagent system, such as poor system performance, excessive processing power requirement, and long delays. The bandwidth and response time performance of the proposed end-to-end system is investigated through the simulation of the proposed agent communication protocol on our private LAN testbed called Hierarchical Agent Network for Intrusion Detection Systems (HAN-IDS). The simulation results show that this system is efficient and extensible since it consumes negligible bandwidth with low cost and low response time on the network.

Recurring traffic congestion is attributable to steadily increasing travel demand coupled with constrained space and financial resources for infrastructure expansion. Another major source of congestion is non-recurrent incidents that disrupt the normal operation of the infrastructure. Aiming to optimize the utilization of the transportation infrastructure, innovative infrastructure management techniques that incorporate on edge technological equipment and information systems need to be adopted to manage recurrent and non-recurrent congestion and reduce their adverse externalities. The framework presented in this thesis lays the foundation for multi-disciplinary semantic web based incident management. During traffic incident response, involved stakeholders will share their knowledge and resources, forming an ad-hoc framework within which each party will focus on its core competencies and cooperate to achieve a coherent incident management process. Negotiation between various response agencies operators is performed using intelligent software agents, alleviating the coordination and synchronization burden of the massive information flow during the incident response. The software agents provide a decision support to human operators based on the reasoning provided from the underlying system knowledge models. Ontological engineering is used to lay the foundation of the knowledge models, which are coded in a web based ontology language, allowing a decentralized access to various elements of the system. The whole system communication infrastructure is based on the Semantic Web technologies. The semantic web facilitates the use of, in an enhanced manner, the already existing web technologies as the communication infrastructure of the proposed system. Its semantic capabilities help to resolve the information and data interoperability issues among various parties. The web services concepts combined with the semantic web allow the direct exploration and access of knowledge models, resources, and data repertories held by various parties. The developed ontology along with the developed software system were tested and evaluated by domain experts and targeted system users. Based on the conducted evaluation, both the ontology and the software system were found to be promising tools in developing pervasive, collaborative and multi-disciplinary traffic incident management systems

In large, open and distributed systems, agents are often used to represent users and act on their behalves. Agents can provide good or bad services or act honestly or dishonestly. Trust and reputation mechanisms are used to distinguish good services from bad ones or honest agents from dishonest ones. My research is focused on trust and reputation management in decentralized systems. Compared with centralized systems, decentralized systems are more difficult and inefficient for agents to find and collect information to build trust and reputation. In this thesis, I propose a Bayesian network-based trust model. It provides a flexible way to present differentiated trust and combine different aspects of trust that can meet agents different needs. As a complementary element, I propose a super-agent based approach that facilitates reputation management in decentralized networks. The idea of allowing super-agents to form interest-based communities further enables flexible reputation management among groups of agents. A reward mechanism creates incentives for super-agents to contribute their resources and to be honest. As a single package, my work is able to promote effective, efficient and flexible trust and reputation management in decentralized systems.

Recurring traffic congestion is attributable to steadily increasing travel demand coupled with constrained space and financial resources for infrastructure expansion. Another major source of congestion is non-recurrent incidents that disrupt the normal operation of the infrastructure. Aiming to optimize the utilization of the transportation infrastructure, innovative infrastructure management techniques that incorporate on edge technological equipment and information systems need to be adopted to manage recurrent and non-recurrent congestion and reduce their adverse externalities. The framework presented in this thesis lays the foundation for multi-disciplinary semantic web based incident management. During traffic incident response, involved stakeholders will share their knowledge and resources, forming an ad-hoc framework within which each party will focus on its core competencies and cooperate to achieve a coherent incident management process. Negotiation between various response agencies operators is performed using intelligent software agents, alleviating the coordination and synchronization burden of the massive information flow during the incident response. The software agents provide a decision support to human operators based on the reasoning provided from the underlying system knowledge models. Ontological engineering is used to lay the foundation of the knowledge models, which are coded in a web based ontology language, allowing a decentralized access to various elements of the system. The whole system communication infrastructure is based on the Semantic Web technologies. The semantic web facilitates the use of, in an enhanced manner, the already existing web technologies as the communication infrastructure of the proposed system. Its semantic capabilities help to resolve the information and data interoperability issues among various parties. The web services concepts combined with the semantic web allow the direct exploration and access of knowledge models, resources, and data repertories held by various parties. The developed ontology along with the developed software system were tested and evaluated by domain experts and targeted system users. Based on the conducted evaluation, both the ontology and the software system were found to be promising tools in developing pervasive, collaborative and multi-disciplinary traffic incident management systems