The traffic domain, and in particular the domain of traffic control, is a highly complex and uncertain domain. A large network of roads, signal controlling systems, vehicles, pedestrians and other traffic units makes the domain intractable. There are great amounts of data available from different parts of traffic, thus there is a need for a method that can take advantage of this data in a systematical manner.In this thesis, we present a prototype Case-based Reasoning (CBR) system which purpose is to execute traffic at a signal controlled pedestrian crossing. The system uses pedestrian- and vehicle data to take decisions in real-time. The system is created as an OSGI bundle and uses the CVIS (Cooperative Vehicle-Infrastructure System) framework to enable communication with other traffic systems and traffic units. myCBR is used as a framework for making the process of retrieving and reusing cases easier. Experts from Norwegian Public Roads Administration were an important resource in defining the structure of the cases and for filling the case base with useful cases. Pedestrian data is obtained by using a Kinect sensor, and the Intention-based Sliding Doors system created by Solem, a previous MSc at our group, is integrated for interpreting the intention of pedestrians at the crossing. Vehicle data is obtained by using simulation software called SCANeR Studio.The results of the project showed that the CBR system adapted to the current traffic situation, and that correct cases were retrieved. These tests were performed in a limited test environment, and to evaluate the system properly, tests in a real environment is necessary.

Contains prestudy within racing data analysis, detailed architecture of software for analysis and implementation details.

The thesis takes a look at the OWASP AppSensor project. The OWASP AppSensor project is about the idea of detecting attacks inside the applicaiton. The thesis compares OWASP AppSensor against both a web application firewall and an intrusion detection system. The comparison is based both on a short litterature study and an experiment performed. The experiment was a set of attacks based on OWASP top ten list which were executed against a simple bank web application. In the experiment the intrusion detection systems, web application firewall and the AppSensor detection points inside the application was tested to see which attacks they where able to detect. The results were quite satisfying for both the web application firewall and AppSensor meanin that they detected many attacks but AppSensors detection was slightly better.

Med sin store vekst, har internett utviklet seg til et lukrativt domene for organisert kriminalitet. Som andre typer organisert kriminalitet er mesteparten av aktiviteten motivert av økonomisk gevinst. I tillegg til økonomisk motiverte trusselen aktører er noen tilsynelatende drevet av politiske motiver, som nasjonalstaters etterretningsorganisasjoner og cyberterrorister. En viktig del av datakriminalitet er å bryte seg inn i datasystemer og sikre fremtidig kontroll over systemene. Når nettkriminelle har klart å få tilgang til et system installerer de ofte et skjult program for å sikre fremtidige tilgang. Dette programmet kalles en bot og rekrutterer den kompromitterte maskinen inn i et botnet. Flere boter under en felles sentral administrasjon kalles et botnet. Denne oppgaven beskriver utformingen av en botnet detektor og rapporterer resultatene fra testing av detektoren på reelle data fra en organisasjon i Norge. Det foreslåtte systemet er designet rundt et klassisk "misuse detection system". Det tar som input nettverksaktivitetslogg som NetFlow, DNS logg og HTTP logg og søker igjennom denne loggen med et stort signaturrett sett sammen av ulike signatursett som deles fritt på internett. Detektoren er basert på fire hovedkomponenter. 1) En algoritme for å kvantifisere risikoen representert ved en signatur, 2) En algoritme for hvitlisting av dårlige signaturer som vil skape falske positiver, 3) En søkemotor for å søke loggfiler med et stort signatursett, og 4) En algoritme for å identifisere kompromittert datamaskiner ved å aggregere alarm data.All komponentene sett under ett ser ut til å gi en betydelig forbedring i forhold til inntrengningsdeteksjon basert på vanlig signatursøk. En av de viktigste forbedringene er at systemene gjør det mye enklere å håndtere dårlige signaturer som skaper mange falske positiver, forbedring synes å være en kombinasjon av hvitlisting av noen av de dårlige signaturene og at fokuset flyttes fra arbeide med alarmer direkte til å jobbe med aggregert klient risikoer.Systemet er komplementære og synergistiske til noen av de nylig foreslåtte system i forskningslitteraturen som Exposure(Bilge, Kirda, Kruegel, & Balduzzi, 2011) og Notos (Antonakakis, Perdisci, Dagon, Lee, & Feamster, 2010)

Modern Graphics Processing Units (GPU) exhibit a high degree of parallelism and over the years have grown to adopt an increasing number of techniques to speed-up photorealistic rendering. One such technique is tessellation, i.e. the recursive subdivision of object elements into finer or coarser parts with the aim of achieving the appropriate amount of detail. The aim of this thesis is an adaptive tessellation algorithm for terrain rendering which can highlight important parts of the terrain while maintaining reasonable performance. This algorithm needs to be crack-free to avoid pixel faults between adjacent patches. A prototype was created in OSG, with GLSL shaders. The adaptive tessellation used three tessellation selection factors based on distance, a tessellation map and normals. An OSG program with five GLSL shaders (vertex, tessellation control, tessellation evaluation, geometry and fragment) was created for the tessellation of the terrain. The greatest advantage of tessellation is the reduced bandwidth between memory and GPU. Tessellation improves FPS, because it's faster to control vertices on the GPU than on the CPU.

To avoid the problem that increasing PHEV demand will further aggravate peak demand hours in the power grid, several different multi-agent scheduling mechanisms have been investigated, including two centralized scheduling mechanisms and two decentralized scheduling mechanisms. For both of the decentralized mechanisms, the PHEV agents choose their own charging plans without relying upon a centralized scheduler, while in the centralized scheduling mechanisms, the PHEVs agents defer control to a central agent for creating their charging plans. From the results, we found that while both the centralized mechanisms and the decentralized mechanisms helped to reduce the average maximum peak, the performance of the centralized mechanisms proved to be highly dependent on how the day-ahead portfolio was calculated. Because of this, the overall best performer was decentralized mechanism, which gave the best results for reducing the average maximum peak without compromising the ability of the PHEVs to charge their batteries too much.

Ever since the invention of EEG have constant attempts been made to give meaning to the oscillating signal recorded. This has resulted in the ability to detect a wide range of different psychological and physiological phenomenon. The goal of this thesis is to be able to recognize different emotions in EEG by the use of a computer and artificial intelligence. Emotions are an especially interesting phenomenon because of the huge impact they have on humans on the daily basis. They constantly guides and modulates our rationality, and is thus in some sense an important part of the definition of human rationality, which again plays an important role in how we behave intelligently and especially how we behave intelligently when interacting with other humans. Machines that interact with humans do however not base their decisions on a rationality that incorporates live human emotions. The effect of this mismatch in rationality between humans and machines results in unwanted behaviour from the machines, and is something most have experienced. The system we propose in this thesis could be used to allow machines to incorporate an interpretation of human emotions in their principles of rationality, in the form of a recognized two-dimensional model of emotions, which could result in a more intelligent interaction with humans. We further restricted our system to the hardware limitations of the commercially available Emotiv EPOC EEG headset, in order to get an indication of the commercial value and general implications of our method. Both unsuccessful and successful systems with similar goals have previously been described in the literature. These systems typically rely on computationally expensive feature extractions, which make them less attractive when a relatively quick response is needed. Moreover, the act of choosing what methods to use in order to extract features entails a degree of subjectivity from the creator, which becomes clear by looking at the share variety of completely different methods used in the different systems. Our system effectively minimizes both of these issues by presenting the signal as it is, expressed in the frequency domain, to an artificial neural network trained by a neuroevolutionary method called HyperNEAT, with promising results. This highlights the importance of using a method that is truly in line with nature of the problem.

It has been shown that most players enjoys playing video games against other human players instead of computer controlled opponents. But most research on artificial intelligence in gaming today focus on just winning in the most effective way possible, instead of making the agents more human-like.Cognitive architectures are designed to emulate how the human brain operates when performing tasks. Very little research has been done on applying cognitive methods to the field of real-time strategy games.In this paper we aim to research the use of a cognitive model that is capable of playing StarCraft: Brood War with decent results. The result is an AI agent implemented with a cognitive framework called LIDA. The resulting agent is only a proof of concept implementation, but we provide suggestions for how it can be improved in the future, and where the problems and limitations of our approach lies.

Evolution in materio aim at revealing computational propertiesin materials. A bottom-up approach combining methods and theoriesfrom evolutionary algorithms, complex systems and unconventionalcomputation, is used to explore inherent computationalproperties of alternative computational materials. This master'sthesis build on earlier and ongoing research at NTNU focusingon interfacing materials. Mecobo is a prototype interface platform,including a physical interface, hardware for stimuli/measurementsand software, developed at NTNU.The Mecobo platform is intended to serve as a platform forexperimenting with unconventional materials' behavior in a dynamiccomplex system. This project describes the existing Mecoboplatform, modications to it and the new extension Cellular AutomataCentral Processing Unit (CA CPU). With the new extension,Mecobo is now capable of modelling dynamic complexsystems with a cellular automata and an unconventional material.The Cellular Automata Central Processing Unit extension is aparallel processor with 32 cores, but supports up to and including64 cores, for controlling each material conguration pin. Thecore is implemented as a stack machine optimized for instructionspace. It uses an 8-bit reduced instruction set computing (RISC)instruction set architecture (ISA).The modied Mecobo platform is tested and conrmed towork with several cellular automata rules and a simple modelconsisting of a theoretical material with constant response.

This thesis presents an approach to controlling Micromanagement in Real-Time Strategy (RTS) computer games using Potential Fields (PF) that are tuned with Multi-Objectve Optimized Evolutionary Algorithms (MOEA), specifically the Nondominated Sorting Genetic Algorithm (NSGA-II). The classic RTS title textit{StarCraft: Broodwar} has been chosen as testing platform due to its status in the competitive AI scene, the amount of detailed information available from previous research and projects, and the free open-source framework Brood War Application Programming Interface (BWAPI). The proposed AI controls its units by placing several types of Potential Fields onto the battlefield. The weights behind the PFs' calculations are optimized using NSGA-II. This work is an attempt to improve on previous methods done with PF in RTS. The results indicate that Multi-Objective Optimization is a suited method for optimizing Potential Fields in RTS games.