Дисертації з теми "Game tree"

Starting with a broad analysis of game story-telling methods, this project seeks to interrogate a player’s occupation and navigation of the possibility space of a game’s meta-narrative and isolate ‘story-telling’ techniques specific and inherent to the computer-game medium. This leads to a series of formalized design schema, on the meta-narrative – game-play relationship, to assist computer-game design practice.

A hybrid methodology of game theory and Monte Carlo Tree Search was developed and the hybrid methodology was tested with various case studies through the nurse scheduling problem to show that it was able to form Pareto front dominance solutions, finding feasible solutions that were optimal and finding feasible partial solutions in over-constrained problems. The performance comparison was carried out with the Genetic Algorithm on the Resident Physician Scheduling problem and showed that the hybrid methodology was able to produce better quality solutions compared to the state of the art approach.

The strategy game Risk is a very popular boardgame, requiring little effort to learn but lots of skill to master.The aim of this project is to explore the fortification phase of thegame, where the player’s troops are moved between territories.Our method is based on adapting Monte Carlo tree search(MCTS) to Risk. To improve the troop movements, we proposetwo techniques, hierarchical search and progressive bias. Thesemethods, combined with other extensions of MCTS are thencompared against a baseline player of the game. Our results showthat hierarchical search improved the MCTS agent’s playingpower and the progressive bias have potential to improve theagent but needs further investigation.
Strategispelet Risk är ett väldigt populärt brädspel som är lätt att lära sig men svårt att bemästra. Syftet med detta projekt är att utforska spelets befästningsfas, där spelarens trupper flyttas mellan territorier. Vår metod är baserad på en anpassning av Monte Carlo trädsökning (MCTS) till Risk. För att förbättra trupprörelserna föreslår vi två tekniker, ”hierarchical search” och ”progressive bias”. Dessa metoder, i kombination med andra tillägg av MCTS, jämförs sedan mot en standard agent i spelet. Våra resultat visar att hierarchical search förbättrade MCTS agentens spelstyrka och att progressivce bias har möjlighet att förbättra agenten men kräver fortsatt utforskning.
Kandidatexjobb i elektroteknik 2020, KTH, Stockholm

Gomoku, also called Five in a row, is one of the earliest checkerboard games invented by humans. For a long time, it has brought countless pleasures to us. We humans, as players, also created a lot of skills in playing it. Scientists normalize and enter these skills into the computer so that the computer knows how to play Gomoku. However, the computer just plays following the pre-entered skills, it doesn’t know how to develop these skills by itself. Inspired by Google’s AlphaGo Zero, in this thesis, by combining the technologies of Monte Carlo Tree Search, Deep Neural Networks, and Reinforcement Learning, we propose a system that trains machine Gomoku players without prior human skills. These are self-evolving players that no prior knowledge is given. They develop their own skills from scratch by themselves. We have run this system for a month and half, during which time 150 different players were generated. The later these players were generated, the stronger abilities they have. During the training, beginning with zero knowledge, these players developed a row-based bottom-up strategy, followed by a column-based bottom-up strategy, and finally, a more flexible and intelligible strategy with a preference to the surrounding squares. Although even the latest players do not have strong capacities and thus couldn’t be regarded as strong AI agents, they still show the abilities to learn from the previous games. Therefore, this thesis proves that it is possible for the machine Gomoku player to evolve by itself without human knowledge. These players are on the right track, with continuous training, they would become better Gomoku players.

This dissertation addresses five fundamental resource allocation problems on networks, all of which have applications to support Homeland Security or industry challenges. In the first application, we model and solve the strategic problem of minimizing the expected loss inflicted by a hostile terrorist organization. An appropriate allocation of certain capability-related, intent-related, vulnerability-related, and consequence-related resources is used to reduce the probabilities of success in the respective attack-related actions, and to ameliorate losses in case of a successful attack. Given the disparate nature of prioritizing capital and material investments by federal, state, local, and private agencies to combat terrorism, our model and accompanying solution procedure represent an innovative, comprehensive, and quantitative approach to coordinate resource allocations from various agencies across the breadth of domains that deal with preventing attacks and mitigating their consequences. Adopting a nested event tree optimization framework, we present a novel formulation for the problem as a specially structured nonconvex factorable program, and develop two branch-and-bound schemes based respectively on utilizing a convex nonlinear relaxation and a linear outer-approximation, both of which are proven to converge to a global optimal solution. We also investigate a fundamental special-case variant for each of these schemes, and design an alternative direct mixed-integer programming model representation for this scenario. Several range reduction, partitioning, and branching strategies are proposed, and extensive computational results are presented to study the efficacy of different compositions of these algorithmic ingredients, including comparisons with the commercial software BARON. The developed set of algorithmic implementation strategies and enhancements are shown to outperform BARON over a set of simulated test instances, where the best proposed methodology produces an average optimality gap of 0.35% (compared to 4.29% for BARON) and reduces the required computational effort by a factor of 33. A sensitivity analysis is also conducted to explore the effect of certain key model parameters, whereupon we demonstrate that the prescribed algorithm can attain significantly tighter optimality gaps with only a near-linear corresponding increase in computational effort. In addition to enabling effective comprehensive resource allocations, this research permits coordinating agencies to conduct quantitative what-if studies on the impact of alternative resourcing priorities. The second application is motivated by the author's experience with the U.S. Army during a tour in Iraq, during which combined operations involving U.S. Army, Iraqi Army, and Iraqi Police forces sought to interdict the transport of selected materials used for the manufacture of specialized types of Improvised Explosive Devices, as well as to interdict the distribution of assembled devices to operatives in the field. In this application, we model and solve the problem of minimizing the maximum flow through a network from a given source node to a terminus node, integrating different forms of superadditive synergy with respect to the effect of resources applied to the arcs in the network. Herein, the superadditive synergy reflects the additional effectiveness of forces conducting combined operations, vis-à-vis unilateral efforts. We examine linear, concave, and general nonconcave superadditive synergistic relationships between resources, and accordingly develop and test effective solution procedures for the underlying nonlinear programs. For the linear case, we formulate an alternative model representation via Fourier-Motzkin elimination that reduces average computational effort by over 40% on a set of randomly generated test instances. This test is followed by extensive analyses of instance parameters to determine their effect on the levels of synergy attained using different specified metrics. For the case of concave synergy relationships, which yields a convex program, we design an inner-linearization procedure that attains solutions on average within 3% of optimality with a reduction in computational effort by a factor of 18 in comparison with the commercial codes SBB and BARON for small- and medium-sized problems; and outperforms these softwares on large-sized problems, where both solvers failed to attain an optimal solution (and often failed to detect a feasible solution) within 1800 CPU seconds. Examining a general nonlinear synergy relationship, we develop solution methods based on outer-linearizations, inner-linearizations, and mixed-integer approximations, and compare these against the commercial software BARON. Considering increased granularities for the outer-linearization and mixed-integer approximations, as well as different implementation variants for both these approaches, we conduct extensive computational experiments to reveal that, whereas both these techniques perform comparably with respect to BARON on small-sized problems, they significantly improve upon the performance for medium- and large-sized problems. Our superlative procedure reduces the computational effort by a factor of 461 for the subset of test problems for which the commercial global optimization software BARON could identify a feasible solution, while also achieving solutions of objective value 0.20% better than BARON. The third application is likewise motivated by the author's military experience in Iraq, both from several instances involving coalition forces attempting to interdict the transport of a kidnapping victim by a sectarian militia as well as, from the opposite perspective, instances involving coalition forces transporting detainees between interment facilities. For this application, we examine the network interdiction problem of minimizing the maximum probability of evasion by an entity traversing a network from a given source to a designated terminus, while incorporating novel forms of superadditive synergy between resources applied to arcs in the network. Our formulations examine either linear or concave (nonlinear) synergy relationships. Conformant with military strategies that frequently involve a combination of overt and covert operations to achieve an operational objective, we also propose an alternative model for sequential overt and covert deployment of subsets of interdiction resources, and conduct theoretical as well as empirical comparative analyses between models for purely overt (with or without synergy) and composite overt-covert strategies to provide insights into absolute and relative threshold criteria for recommended resource utilization. In contrast to existing static models, in a fourth application, we present a novel dynamic network interdiction model that improves realism by accounting for interactions between an interdictor deploying resources on arcs in a digraph and an evader traversing the network from a designated source to a known terminus, wherein the agents may modify strategies in selected subsequent periods according to respective decision and implementation cycles. We further enhance the realism of our model by considering a multi-component objective function, wherein the interdictor seeks to minimize the maximum value of a regret function that consists of the evader's net flow from the source to the terminus; the interdictor's procurement, deployment, and redeployment costs; and penalties incurred by the evader for misperceptions as to the interdicted state of the network. For the resulting minimax model, we use duality to develop a reformulation that facilitates a direct solution procedure using the commercial software BARON, and examine certain related stability and convergence issues. We demonstrate cases for convergence to a stable equilibrium of strategies for problem structures having a unique solution to minimize the maximum evader flow, as well as convergence to a region of bounded oscillation for structures yielding alternative interdictor strategies that minimize the maximum evader flow. We also provide insights into the computational performance of BARON for these two problem structures, yielding useful guidelines for other research involving similar non-convex optimization problems. For the fifth application, we examine the problem of apportioning railcars to car manufacturers and railroads participating in a pooling agreement for shipping automobiles, given a dynamically determined total fleet size. This study is motivated by the existence of such a consortium of automobile manufacturers and railroads, for which the collaborative fleet sizing and efforts to equitably allocate railcars amongst the participants are currently orchestrated by the \textit{TTX Company} in Chicago, Illinois. In our study, we first demonstrate potential inequities in the industry standard resulting either from failing to address disconnected transportation network components separately, or from utilizing the current manufacturer allocation technique that is based on average nodal empty transit time estimates. We next propose and illustrate four alternative schemes to apportion railcars to manufacturers, respectively based on total transit time that accounts for queuing; two marginal cost-induced methods; and a Shapley value approach. We also provide a game-theoretic insight into the existing procedure for apportioning railcars to railroads, and develop an alternative railroad allocation scheme based on capital plus operating costs. Extensive computational results are presented for the ten combinations of current and proposed allocation techniques for automobile manufacturers and railroads, using realistic instances derived from representative data of the current business environment. We conclude with recommendations for adopting an appropriate apportionment methodology for implementation by the industry.
Ph. D.

Artificial Intelligence (AI) is used in many games and quite often the Non-Player Character(NPC)s simulate humans. To make the human NPCs believable and feel alive they need to be as human-like as possible in their behaviour. Three features commonly used to make an NPC human-like are needs, like eating or sleeping, social relationships and personalities. The objective of this thesis was to create an AI with different personalities that the NPCs may have in the game Future Happiness Challenge (FHC) and compare these personalities as well as try to find out whether personalities enhance the game experience. Three different personalities are implemented; Selfish, Selfless and Balanced. These are used as FHC presents the option to play either selfish or selfless. It can be played as a team or as an individualist that does not care about the others. This thesis tries to answer the question whether a player prefers a selfless NPC to a selfish in a game where this option is available. These extremes are also compared to a balanced NPC. When implementing the AI and the personalities, a Behaviour Tree (BT) was used and the main features of this implementation is presented to give an example of how personalities can be implemented in a game like FHC. The results suggest that personalities enhance the game experience and an interesting correlation can be seen between the players preference of an NPC and their own personality when playing the game. Rather than always preferring a selfish or a selfless NPC, the players seem to prefer the NPC which has a personality closely related to their own within the game.
Artificiell intelligens (AI) används i många spel och det är vanligt att datorstyrda karaktärer (Non-Player Characters) föreställer människor. För att göra dessa karaktärer trovärdiga och ge känslan av att leva så behöver deras beteenden göras så mänskliga som möjligt. Tre egenskaper som ofta används för att göra karaktärerna mänskliga är behov, såsom att äta eller sova, sociala relationer och personligheter. Målet med det här arbetet var att skapa en AI med olika personligheter som karaktärerna kan ha i spelet Future Happiness Challenge (FHC) och jämföra dessa personligheter samt försöka ta reda på om personligheter ökar spelupplevelsen. Tre olika personligheter implementerades; självisk, osjälvisk och balanserad. Dessa valdes då FHC ger spelaren möjligheten att spela antingen själviskt eller osjälviskt. Det kan spelas som ett lag eller som en egoist som inte bryr sig om de andra. Denna rapport försöker svara på frågan om spelaren föredrar en osjälvisk karaktär över en självisk i ett spel där denna möjlighet finns. Dessa extrema personligheter jämförs också med en balanserad. Under utvecklingen av AIn och personligheterna användes tekniken Behaviour Tree (BT) och större delen av implementationen är presenterad i detta arbete för att ge ett exempel på hur personligheter kan implementeras i ett spel som FHC. Resultaten föreslår att personligheter ökar spelupplevelsen och ett intressant samband kan ses mellan spelarnas preferens av NPC och spelarnas personligheter i FHC. Istället för att alltid föredra en självisk eller osjälvisk NPC, verkar spelarna föredra den NPC som har en personlighet lik sin egen i spelet.

The purpose of this thesis is to evaluate the possibility of predicting difficulty, measured in average human success rate (AHSR), across game levels of a mobile game using a general AI algorithm. We implemented and tested a simulation based bot using MCTS for Candy. Our results indicate that AHSR can be predicted accurately using MCTS, which in turn suggests that our bot could be used to streamline game level development. Our work is relevant to the field of AI, especially the subfields of MCTS and single-player stochastic games as Candy, with its diverse set of features, proved an excellent new challenge for testing the general capabilities of MCTS. The results will also be valuable to companies interested in using AI for automatic testing of software.

The subjects of perfect information games, machine learning and computational intelligence combine in an experiment that investigates a method to build the skill of a game-playing agent from zero game knowledge. The skill of a playing agent is determined by two aspects, the first is the quantity and quality of the knowledge it uses and the second aspect is its search capacity. This thesis introduces a novel representation language that combines symbols and numeric elements to capture game knowledge. Insofar search is concerned; an extension to an existing knowledge-based search method is developed. Empirical tests show an improvement over alpha-beta, especially in learning conditions where the knowledge may be weak. Current machine learning techniques as applied to game agents is reviewed. From these techniques a learning framework is established. The data-mining algorithm, ID3, and the computational intelligence technique, Particle Swarm Optimisation (PSO), form the key learning components of this framework. The classification trees produced by ID3 are subjected to new post-pruning processes specifically defined for the mentioned representation language. Different combinations of these pruning processes are tested and a dominant combination is chosen for use in the learning framework. As an extension to PSO, tournaments are introduced as a relative fitness function. A variety of alternative tournament methods are described and some experiments are conducted to evaluate these. The final design decisions are incorporated into the learning frame-work configuration, and learning experiments are conducted on Checkers and some variations of Checkers. These experiments show that learning has occurred, but also highlights the need for further development and experimentation. Some ideas in this regard conclude the thesis.
Dissertation (MSc)--University of Pretoria, 2007.
Computer Science
MSc
Unrestricted

The majority of terrestrial plants associate with fungi in symbiotic resource-exchange relationships called mycorrhizae. Mycorrhizal networks (MNs) arise when the same fungus is connected to multiple plants, allowing for interplant resource transfer and impacting ecosystem functions. Recent work suggests MNs also transfer defense-related information from pathogen-, herbivore-, or mechanically-damaged plants to unharmed neighbors. I investigated the defense pathways involved in defense-related signal transfer in ectomycorrhizal systems. Paired Douglas-fir seedlings were grown with varying levels of belowground connectivity (soil water only; soil water and MNs; soil water, MNs, and roots), and a defense response was stimulated in donor seedlings by methyl jasmonate. After 24 and 48 hrs, I measured expression of two regulatory genes on the jasmonate and ethylene pathways. Receiver response was unrelated to hormone treatment of donors in either gene, but the jasmonate response of donor and receiver pairs was correlated across treatments. Positive expression of both genes across donors and receivers and pervasive presence of spider mites suggested signal transfer may either have not occurred or been masked by already ongoing defensive responses. Results indicate the complexity of these systems, and further work is needed to better characterize defense signal transfer via ectomycorrhizal networks. Because of the importance of these mycorrhizal systems to ecosystem functioning, it is crucial that resource managers and scientists have a good understanding of mycorrhizal ecology. However, lower student interest in plants and fungi combined with difficulties visualizing belowground processes present challenges for teaching and learning mycorrhizal concepts. To address this, I co-created the digital plant-centric action-based game Shroomroot for use in lower level postsecondary settings. I conducted a pre-test/post-test evaluation of Shroomroot in a 2nd year postsecondary Introduction to Soil Science course. Students’ knowledge of mycorrhizal ecology increased after playing Shroomroot, and engagement with mycorrhizal content tended to increase after gameplay. These exploratory results suggest positive potential for action-based plant-oriented digital games in the higher education classrooms. Both studies focus on improving our understanding of mycorrhizae and mycorrhizal networks, ecologically and pedagogically. Greater understanding of mycorrhizae has the potential to improve our multi-faceted relationships with the ecosystems upon which we depend.
Forestry, Faculty of
Forest and Conservation Sciences, Department of
Graduate

This study explores different actors’ perceptions on how the democratization process is being influenced in Kosovo. It is based upon empirical facts gathered during a field study in Kosovo. The material was mainly gathered through interviews with different representatives from international organizations present in Kosovo and with members of the Kosovar political elite.

The theoretical standpoint in the study draws on the theory of polyarchy for defining the concept of democracy, and the two-level game. The interaction between domestic and international actors within the democratization process is highlighted.

The conclusions are that different actors are influencing the ongoing democratization process in Kosovo, mainly international external ones acting in several ways and on different levels. This influence is considered by the international community and the domestic political elite to be necessary both for the continued development of democracy and for the preservation of stability. However as the influence has more or less become institutionalized, problems occur when responsibility is handed over to domestic authorities without accountability being made clear.

We consider a class of problems characterized by the need for a string based identifier that reflect the ontology of the application domain. We present rules for string-based identifier schemas that facilitate fast filtering in databases used for this class of problems. We provide runtime analysis of our schema and experimentally compare it with another solution. We also discuss performance in our solution to a game engine. The string-based identifier schema can be used in addition scenarios such as cloud computing. An identifier schema adds metadata about an element. So the solution hinges on additional memory but as long as queries operate only on the included metadata there is no need to load the element from disk which leads to huge performance gains.

It is well known that the branching factor of a computer based board game has an effect on how long a searching AI algorithm takes to search through the game tree of the game. Something that is not as known is that the branching factor may have an additional effect for certain types of AI algorithms. The aim of this work is to evaluate if the win rate of an evolutionary AI algorithm is affected by the branching factor of the board game it is applied to. To do that, an experiment is performed where an evolutionary algorithm known as “Genetic Minimax” is evaluated for the two low branching factor board games Othello and Gomoku (Gomoku is also known as 5 in a row). The performance here is defined as how many times the algorithm manages to win against another algorithm. The results from this experiment showed both some promising data, and some data which could not be as easily interpreted. For the game Othello the hypothesis about this particular evolutionary algorithm appears to be valid, while for the game Gomoku the results were somewhat inconclusive. For the game Othello the performance of the genetic minimax algorithm was comparable to the alpha-beta algorithm it played against up to and including depth 4 in the game tree. After that however, the performance started to decline more and more the deeper the algorithms searched. The branching factor of the game may be an indirect cause of this behaviour, due to the fact that as the depth increases, the search space increases proportionally to the branching factor. This increase in the search space due to the increased depth, in combination with the settings used by the genetic minimax algorithm, may have been the cause of the performance decline after that point.

Exploiter des jeux de données labelisés est très utile, non seulement pour entrainer des modèles et mettre en place des procédures d'analyses prédictives, mais aussi pour améliorer la compréhension d'un domaine. La découverte de sous-groupes a été l'objet de recherches depuis deux décennies. Elle consiste en la découverte de règles couvrants des ensembles d'objets ayant des propriétés intéressantes, qui caractérisent une classe cible donnée. Bien que de nombreux algorithmes de découverte de sous-groupes aient été proposés à la fois dans le cas des données transactionnelles et numériques, la découverte de règles dans des données séquentielles labelisées a été bien moins étudiée. Dans ce contexte, les stratégies d'exploration exhaustives ne sont pas applicables à des cas d'application rééls, nous devons donc nous concentrer sur des approches heuristiques. Dans cette thèse, nous proposons d'appliquer des modèles de bandit manchot ainsi que la recherche arborescente de Monte Carlo à l'exploration de l'espace de recherche des règles possibles, en utilisant un compromis exploration-exploitation, sur différents types de données tels que les sequences d'ensembles d'éléments, ou les séries temporelles. Pour un budget temps donné, ces approches trouvent un ensemble des top-k règles decouvertes, vis-à-vis de la mesure de qualité choisie. De plus, elles ne nécessitent qu'une configuration légère, et sont indépendantes de la mesure de qualité utilisée. A notre connaissance, il s'agit de la première application de la recherche arborescente de Monte Carlo au cas de la fouille de données séquentielles labelisées. Nous avons conduit des études appronfondies sur différents jeux de données pour illustrer leurs plus-values, et discuté leur résultats quantitatifs et qualitatifs. Afin de valider le bon fonctionnement d'un de nos algorithmes, nous proposons un cas d'utilisation d'analyse de jeux vidéos, plus précisémment de matchs de Rocket League. La decouverte de règles intéressantes dans les séquences d'actions effectuées par les joueurs et leur exploitation dans un modèle de classification supervisée montre l'efficacité et la pertinence de notre approche dans le contexte difficile et réaliste des données séquentielles de hautes dimensions. Elle permet la découverte automatique de techniques de jeu, et peut être utilisée afin de créer de nouveaux modes de jeu, d'améliorer le système de classement, d'assister les commentateurs de "e-sport", ou de mieux analyser l'équipe adverse en amont, par exemple
It is extremely useful to exploit labeled datasets not only to learn models and perform predictive analytics but also to improve our understanding of a domain and its available targeted classes. The subgroup discovery task has been considered for more than two decades. It concerns the discovery of rules covering sets of objects having interesting properties, e.g., they characterize a given target class. Though many subgroup discovery algorithms have been proposed for both transactional and numerical data, discovering rules within labeled sequential data has been much less studied. In that context, exhaustive exploration strategies can not be used for real-life applications and we have to look for heuristic approaches. In this thesis, we propose to apply bandit models and Monte Carlo Tree Search to explore the search space of possible rules using an exploration-exploitation trade-off, on different data types such as sequences of itemset or time series. For a given budget, they find a collection of top-k best rules in the search space w.r.t chosen quality measure. They require a light configuration and are independent from the quality measure used for pattern scoring. To the best of our knowledge, this is the first time that the Monte Carlo Tree Search framework has been exploited in a sequential data mining setting. We have conducted thorough and comprehensive evaluations of our algorithms on several datasets to illustrate their added-value, and we discuss their qualitative and quantitative results. To assess the added-value of one or our algorithms, we propose a use case of game analytics, more precisely Rocket League match analysis. Discovering interesting rules in sequences of actions performed by players and using them in a supervised classification model shows the efficiency and the relevance of our approach in the difficult and realistic context of high dimensional data. It supports the automatic discovery of skills and it can be used to create new game modes, to improve the ranking system, to help e-sport commentators, or to better analyse opponent teams, for example

Ett spels karta är begränsande i det att när man valt en viss storlek kan man inte gå utanför den ramen utan att göra relativt resurskrävande operationer. Denna undersökning genomfördes för att se om en trädstruktur kan användas som lösning att hantera en honeycomb-struktur på ett lämpligt sätt för att ständigt kunna utöka ett spels karta. Resultatet visar att det är möjligt att använda trädstrukturen relativt bra till växande kartor men att det inte är att rekommendera till spel, eftersom strukturen i sig är en omväg. I samband med denna karta skapades en slumpmässig path-generator som skulle kunna användas till att generellt skapa slumpmässiga kartor i spel. För att se vad försvårigheter man stöter på när man utvecklar en slumpad map-generator, vilket visade sig vara svårt i och med att man hittade många specialfall.

Abstract. Autonomous agents, also known as AI agents, are staples in modern video games. They take a lot of roles, everything from being quest-givers in roleplaying games, to opposing forces in action- and shooter games. Crafting an AI that is not only easy to create, but also retains humanlike and believable behavior, has always represented a challenge to the development industry, and has in several cases ended up with open world games using AI systems that limit the AI agents to simple moving patterns. In this thesis, a form of AI systems more commonly used in simulation games such as The Sims video game series, are taken and implemented in an environment that could possibly be seen in an open world game. After the implementation, a set of tests were performed on a group of testers which resulted in the insight that a majority of the testers, when asked to compare their experience to other games, found this implementation to feel more lifelike and realistic.

Dans cette thèse, nous avons étudié les problèmes de décisions séquentielles, avec comme application la gestion de stocks d'énergie. Traditionnellement, ces problèmes sont résolus par programmation dynamique stochastique. Mais la grande dimension, et la non convexité du problème, amènent à faire des simplifications sur le modèle pour pouvoir faire fonctionner ces méthodes. Nous avons donc étudié une méthode alternative, qui ne requiert pas de simplifications du modèle: Monte Carlo Tree Search (MCTS). Nous avons commencé par étendre le MCTS classique (qui s'applique aux domaines finis et déterministes) aux domaines continus et stochastiques. Pour cela, nous avons utilisé la méthode de Double Progressive Widening (DPW), qui permet de gérer le ratio entre largeur et profondeur de l'arbre, à l'aide de deux méta paramètres. Nous avons aussi proposé une heuristique nommée Blind Value (BV) pour améliorer la recherche de nouvelles actions, en utilisant l'information donnée par les simulations passées. D'autre part, nous avons étendu l'heuristique RAVE aux domaines continus. Enfin, nous avons proposé deux nouvelles méthodes pour faire remonter l'information dans l'arbre, qui ont beaucoup amélioré la vitesse de convergence sur deux cas tests. Une part importante de notre travail a été de proposer une façon de mêler MCTS avec des heuristiques rapides pré-existantes. C'est une idée particulièrement intéressante dans le cas de la gestion d'énergie, car ces problèmes sont pour le moment résolus de manière approchée. Nous avons montré comment utiliser Direct Policy Search (DPS) pour rechercher une politique par défaut efficace, qui est ensuite utilisée à l'intérieur de MCTS. Les résultats expérimentaux sont très encourageants. Nous avons aussi appliqué MCTS à des processus markoviens partiellement observables (POMDP), avec comme exemple le jeu de démineur. Dans ce cas, les algorithmes actuels ne sont pas optimaux, et notre approche l'est, en transformant le POMDP en MDP, par un changement de vecteur d'état. Enfin, nous avons utilisé MCTS dans un cadre de méta-bandit, pour résoudre des problèmes d'investissement. Le choix d'investissement est fait par des algorithmes de bandits à bras multiples, tandis que l'évaluation de chaque bras est faite par MCTS. Une des conclusions importantes de ces travaux est que MCTS en continu a besoin de très peu d'hypothèses (uniquement un modèle génératif du problème), converge vers l'optimum, et peut facilement améliorer des méthodes suboptimales existantes.

This thesis deals with general game playing agent systems. On the contrary with common agents, which are designed only for a specified task or a game, general game playing agents have to be able to play basically any arbitrary game described in a formal declarative language. The biggest challenge is that the game rules are not known beforehand, which makes it impossible to use some optimizations or to make a good heuristic function. The thesis consists of a theoretical and a practical part. The first part introduces the field of general game playing agents, defines the Game Description Language and covers construction of heuristic evaluation functions and their integration within the Monte Carlo tree search algorithm. In the practical part, a general method of creating a new heuristic function is presented, which is later integrated into a proper agent, which is compared then with other systems.

This thesis presents the implementation and functionality of the user interface for creating behavior trees in the Unreal Engine (version 4.10). The thesis analyzes the final version of the behavior trees in a game development project carried out over one year with a group ranging between four and seven members. The game which is analyzed is a third person adventure game which contains four types of simple behavior trees. These include two enemies who mainly move towards the player to attack whenever in range and two bosses with individual behavior. The thesis describes the various types of nodes available in the Unreal Engine as well as how the behavior trees in the game are structured. Focus is placed on how the structure achieves the required result and how the process resulted in the final version of the behavior trees.
Detta examensarbete beskriver implementationen och funktionaliteten av användargränssnittet för att skapa beteendeträd i Unreal Engine (version 4.10). Arbetet analyserar den slutgiltiga versionen av beteendeträden i ett spelutvecklingsprojekt som utfördes under ett år med en grupp vars antal växlade mellan fyra och sju medlemmar. Spelat som analyseras är ett tredjepersons äventyrsspel som innehåller fyra typer av grundläggande beteendeträd. Två fiender som huvudsakligen rör sig mot spelaren för att anfalla när de är inom räckhåll, och två bossar med individuella beteenden. Arbetet beskriver de olika typerna av noder tillgängliga i Unreal Engine och även hur beteendeträden i spelet är uppbyggda. Fokus läggs på hur strukturen uppnår det nödvändiga resultatet samt på hur processen resulterade i den slutgiltiga versionen av beteendeträden.

The goal of this master's thesis is to create an artificial intelligence for the Bang! game. There is a full description of the Bang! game, it's entire rules, player's using strategy principles and game analysis from UI point of view included. The thesis also resumes methods of the artificial intelligence and summarizes basic information about the domain of game theory. Next part describes way of the implementation in C++ language and it's proceeding with use of Bayes classification and decision trees based on expert systems. Last part represent analysis of altogether positive results and the conclusion with possible further extensions.

The thesis is focused on solving the Sokoban game using artificial intelligence algorithms. The first part of the thesis describes the Sokoban game, state space and selected state space search methods. In the second part selected methods were implemented and graphic user interface was created in the Python environment. Comparative experiments were executed in the final part.

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO
COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR
PROGRAMA DE EXCELENCIA ACADEMICA
TREG é um jogo para o treinamento em Engenharia de Requisitos, especificamente na técnica de Workshops. Foi criado no Second Life, usando suas possibilidades de modelagem 3D e programação. Este trabalho apresenta uma experiência de exploração num processo de prototipação e técnicas para o desenvolvimento do jogo. O ciclo de desenvolvimento do TREG utiliza o processo de prototipação baseado no livro Effective Prototyping for Software Makers, que descreve um processo iterativo que foi customizado para a criação do jogo educacional neste mundo virtual. Branching Stories é um género de simulações que dá uma visão geral de todos os caminhos possíveis do jogador no TREG. A técnica de Cenários é utilizada para a especificação das simulações modeladas no gráfico de Branching Stories. O desenho do TREG utiliza especificações feitas usando a técnica de Cenários para a modelagem de diferentes perspectivas de software. As maquinas de Estados modelam o comportamento dinâmico dos elementos de software, o diagrama de classes representa os objetos criados no Second Life e o diagrama de comunicação mostra as relações e a passagem de mensagens entre os objetos. Além disso, este trabalho apresenta os resultados dos testes de usabilidade e as dificuldades apresentadas pelos participantes para a utilização desta ferramenta.
TREG is a game for training in Requirements Engineering, specifically in the Workshops technique. It was created in Second Life using its building and scripting possibilities. This works presents an exploration in the use of a prototyping process and techniques for developing the game. The prototyping process of the book Effective Prototyping for Software Makers is an iterative process which was customized for the development of the game in this virtual world. Branching Stories is a simulation genre that gives an overview of all the possible paths the player can take in TREG. Scenario is a Requirements Engineering technique used for the specification of the simulations modeled in the Branching Stories graph. The design of TREG used the Scenarios specifications for modeling the software perspectives. The state machine diagrams shows the dynamic behavior of the TREG objects, a class diagram represents the objects created in Second Life and the communication diagrams shows the relationships and the flow of messages between objects. This work also shows the results of the evaluation made and the difficulties presented by the participants whereas TREG was used.
TREG es un juego para el entrenamiento en Ingeniería de Requisitos específicamente en la técnica de Workshops. Fue creado en Second Life, usando las posibilidades de modelamiento 3D y programación en este ambiente. Este trabajo presenta una experiencia de exploración en un proceso de prototipación y técnicas para desarrollar el juego. El ciclo de desarrollo de TREG utiliza el proceso de prototipación basado en el libro Effective Prototyping for Software Makers, este es un proceso iterativo que fue personalizado para la creación del juego educacional en este mundo virtual. Branching Stories es un género de simulaciones que da una visión global de todos los caminos posibles del jugador en TREG. La técnica de Escenarios es usada para la especificación de las simulaciones modeladas en el grafo de Branching Stories. El diseno de TREG utiliza las especificaciones realizadas con la técnica de escenarios para modelar diversas perspectivas del software. Las Maquinas de Estados modelan el comportamiento dinámico de los elementos del software, el diagrama de clases representa los objetos creados en Second Life y el diagrama de comunicación muestra las relaciones y paso de mensajes entre los objetos. Además, este trabajo muestra los resultados de la evaluación realizada así como las dificultades presentadas por los participantes al utilizar esta herramienta.

The increasing need for more energy in the world is making the oil companies search for hydrocarbons in deeper water and with longer distances between the well and platform(step-out) never experienced before. Current systems are designed to operate in deepwater with relatively short step-outs. Since the search now is starting to exceed this, the operational conditions are changed and the electro-hydraulic system is closing in on its operational limits. Hence, the increasing need for a system that supports the long step-outs and the deeper water. Aker Kvaerner has started looking into the development for an All Electric Subsea Production system. In co-operation with Aker Kvaerner Subsea, a functional design specification for an electric subsea gate valve actuator has been developed and included in this thesis. This functional design specification will include all the requirements that a subsea electric actuator must conform to. This includes the operational requirements, size limitations, failsafe needs etc. Based on this specification, a design for a electric actuator and including motor driver control system layout on the all electric x-mas tree is proposed as a part of the All Electric Subsea Production system. The actuator will be used to operate the gate valves on a x-mas tree which controls the flow of hydrocarbons through the valves into the pipelines. Most of the existing subsea gate valve actuators are currently using hydraulic oil to operate the valves linearly. Here a concept is being suggested in replacement of the hydraulic fluid, while still complying with the necessary design specification. A number of electrical machines for potential use in this application are presented and evaluated. A Permanent Magnet Synchronous Machine is being investigated further. These machines can be custom made, and can also be made with built in redundancy. Calculation of size for the Permanent Magnet Synchronous Machine has been done, to determine suitability within the space limitation of the apparatus. Two layouts for controlling the actuators on the all electric x-mas tree are proposed. These two are also evaluated against the standards made for subsea production systems and the no 'single point failure' requirement made by the industry. The chosen solution proposed for controlling the electrical actuators is an integrated design, where all actuators have their own motor driver inside. This thesis conclude that an electrical actuator will be technically and mechanically possible to build(from an electrical perspective), using the design and control layout proposed.

Monte Carlo Tree Search (MCTS) is an AI technique that has been successfully applied to many deterministic games of perfect information, leading to large advances in a number of domains, such as Go and General Game Playing. Imperfect information games are less well studied in the field of AI despite being popular and of significant commercial interest, for example in the case of computer and mobile adaptations of turn based board and card games. This is largely because hidden information and uncertainty leads to a large increase in complexity compared to perfect information games. In this thesis MCTS is extended to games with hidden information and uncertainty through the introduction of the Information Set MCTS (ISMCTS) family of algorithms. It is demonstrated that ISMCTS can handle hidden information and uncertainty in a variety of complex board and card games. This is achieved whilst preserving the general applicability of MCTS and using computational budgets appropriate for use in a commercial game. The ISMCTS algorithm is shown to outperform the existing approach of Perfect Information Monte Carlo (PIMC) search. Additionally it is shown that ISMCTS can be used to solve two known issues with PIMC search, namely strategy fusion and non-locality. ISMCTS has been integrated into a commercial game, Spades by AI Factory, with over 2.5 million downloads. The Information Capture And ReUSe (ICARUS) framework is also introduced in this thesis. The ICARUS framework generalises MCTS enhancements in terms of information capture (from MCTS simulations) and reuse (to improve MCTS tree and simulation policies). The ICARUS framework is used to express existing enhancements, to provide a tool to design new ones, and to rigorously define how MCTS enhancements can be combined. The ICARUS framework is tested across a wide variety of games.

La thèse porte sur les deux thèmes des Jeux combinatoires et de la théorie des graphes. Elle est divisée en deux parties.1) Le jeu de Domination et ses variantes: Il s'agit d'un jeu combinatoire qui consiste à marquer les sommets d'un graphe de telle sorte qu'un sommet marqué n'ait aucun voisin marqué. Le joueur marquant le dernier sommet est déclaré gagnant. Le calcul des stratégies gagnantes étant NP-difficile pour un graphe quelconque, nous avons étudié des familles particulières de graphes comme les chemins, les scies ou les chenilles. Pour ces familles on peut savoir en temps polynomial si un graphe est perdant. Nous avons également étudié 28 variantes du jeu de domination, dont les 12 variantes définies par J. Conway sur un jeu combinatoire quelconque. 2) Le nombre chromatique ludique des arbres: Ce paramètre est calculé à partir d'un jeu de coloration où Alice et Bob colorient alternativement et proprement un sommet d'un graphe G avec l'une des k couleurs. L'objectif d'Alice est de colorier complètement le graphe alors que Bob doit l'en empêcher. Nous nous sommes intéressés au jeu avec 3 couleurs sur un arbre T. Nous souhaitons déterminer les arbres ayant un nombre chromatique ludique 3, soit ceux pour lesquels Alice a une stratégie gagnante avec 3 couleurs. Ce problème semblant difficile à résoudre sur les arbres, nous avons résolu des sous-familles: les 1-chenilles puis les chenilles sans trous
Part 1: Domination Game and its variantsDomination game is a combinatorial game that consists in marking vertices of a graph so that a marked vertex has no marked neighbors. The first player unable to mark a vertex loses the game.Since the computing of winning strategies is an NP-hard problem for any graphs, we examine some specific families of graphs such as complete k-partite graphs, paths or saws. For these families, we establish the set of losing elements. For other families, such as caterpillars, we prove that exists a polynomial algorithm for the computation of outcome and winning strategies. No polynomial algorithm has been found to date for more general families, such as trees.We also study 28 variants of Domination game, including the 12 variants defined by J. Conway for any combinatorial game. Using game functions, we find the set of losing paths for 10 of these 12 variants. We also investigate 16 variants called diameter, for instance when rules require to play on the component that has the largest diameter.Part 2: The game chromatic number of treesThis parameter is computed from a coloring game: Alice and Bob alternatively color the vertices of a graph G, using one of the k colors in the color set. Alice has to achieve the coloring of the entire graph whereas Bob has to prevent this. Faigle and al. proved that the game chromatic number of a tree is at most 4. We undertake characterization of trees with a game chromatic number of 3. Since this problem seems difficult for general trees, we focus on sub-families: 1-caterpillars and caterpillars without holes.For these families we provide the characterization and also compute winning strategies for Alice and Bob. In order to do so, we are led to define a new notion, the bitype, that for a partially-colored graph G associates two letters indicating who has a winning strategy respectively on G and G with an isolated vertex. Bitypes allow us to demonstrate several properties, in particular to compute the game chromatic number of a graph from the bitypes of its connected components

In a paper by Beeler and Hoilman, the traditional game of peg solitaire is generalized to graphs in the combinatorial sense. One of the important open problems in this paper was to classify solvable trees. In this thesis, we will give necessary and sufficient conditions for the solvability for all trees with diameter four. We also give the maximum number of pegs that can be left on such a graph under the restriction that we jump whenever possible.

The concept of evolving components of an artificial intelligence (AI) has seen increased interest in recent years as the power and complexity of AI has grown. In entertainment software, this AI can impact the player's experiences and enjoyment through elements such as the level of difficulty of the player's competition. Therefore AI development is an important research topic, especially as development is considered difficult by the video game industry. This work applies the evolutionary computing paradigm to a turn-based domain by evolving team strategies. These strategies are represented as behaviour trees, a formalism found in the video game industry and well-suited to the evolutionary algorithm due to their flexibility and tree structure. During the evolutionary process, strategies are evaluated in Battle for Wesnoth, an open-source game with a stochastic nature. A fitness function is defined to assign strategies a numerical strength value, along with a second performance metric that is robust to the variance found in the domain. The evolutionary algorithm then recombines strategies with high strength values, using evolutionary operators from the literature such as crossover and mutation. Later experiments focus on evolutionary algorithm parameters, including comparing a variety of fitness functions to provide insights into their use. Starting from a number of initial states, numerous strategies are evolved using this algorithm. These initial states are a null strategy, randomly-generated strategies, and a number of hand-built strategies. The evolved strategies are then evaluated in-game, and will be shown to be measurably stronger than the initial strategies.
L'application de l'informatique évolutive au domaine de l'Intelligence Artificielle (IA) émerge naturellement. Pour l'industrie du divertissement (jeux ludiques), IA fait référence aux différents éléments visuels ou non qui déterminent une partie importante de l'expérience du joueur. Cette responsabilité fait en sorte que développer des système d'IA n'est pas une tâche simple, ce qui en fait un domaine de recherche intéressant. Ce travail applique les concepts d'informatique évolutive a un jeu par tour en lassant évoluer des stratégies d'IA. Les stratégies sont représentées par des arbres de comportement. Nous retrouvons cet automate dans l'industrie des jeux digitales, en général il est flexibles et bien adapté a l'informatique évolutive. Durant le processus d'évolution les stratégies sont évaluées dans le jeu Battle for Wesnoth, un jeu logiciel libre de nature stochastique. Une fonction d'aptitude est assignée aux stratégies afin de déterminer numériquement leur efficacité, de plus l'algorithme utilise des mesures robuste à la variance du processus. L'algorithme d'évolution recombine les stratégies efficaces utilisant des opérateurs évolutionnaires basés sur littérature par exemple, mutation ou transition. De plus amples expériences se concentrent sur différents paramètres et leurs utilités, par exemple les fonctions d'aptitudes. L'algorithme utilise de multiples états initiales : stratégie nulle, stratégie issue du hasard et stratégies de conception humaine. Les stratégies sont ensuite testées dans le jeu. Il est démontré par la suite que les stratégies évoluées par l'algorithme sont plus efficace que leurs états initiaux.

Efficient best-first search algorithms have been developed for deterministic two-player games with two-outcome.We present a formal framework to represent such best-first search algorithms.The framework is general enough to express popular algorithms such as Proof Number Search, Monte Carlo Tree Search, and the Product Propagation algorithm.We then show how a similar framework can be devised for two more general settings: two-player games with multiple outcomes, and the model checking problem in modal logic K.This gives rise to new Proof Number and Monte Carlo inspired search algorithms for these settings.Similarly, the alpha-beta pruning technique is known to be very important in games with sequential actions.We propose an extension of this technique for stacked-matrix games, a generalization of zero-sum perfect information two-player games that allows simultaneous moves.

This thesis presents an architecture for an agent that can play the real-time strategy game Starcraft 2 (SC2) by applying Monte Carlo Tree Search (MCTS) together with genetic algorithms and machine learning methods. Together with the MCTS search, a light-weight and accurate combat simulator for SC2 as well as a build order optimizer are presented as independent modules. While MCTS has been well studied for turn-based games such as Go and Chess, its performance has so far been less explored in the context of real-time games. Using machine learning and planning methods in real-time strategy games without requiring long training times has proven to be a challenge. This thesis explores how a model based approach, based on the rules of the game, can be used to achieve a well performing agent.
Denna uppsats presenterar en arkitektur för ett program som kan spela realtidsspelet Starcraft 2 (SC2) genom att använda Monte Carlo Tree Search (MCTS) tillsammans med genetiska algoritmer och maskininlärningsmetoder. Tillsammans med MCTS-sökningen så presenteras också en snabb och exakt stridssimulator för SC2 samt en optimeringsalgoritm för bygg-ordningar som separata moduler. MCTS has studerats mycket inom turordningsbaserade spel som till exempel Go och Schack, däremot så har det utforskats mindre när det kommer till realtidsspel. Att använda maskininlärning och planeringsalgoritmer i realtidsstrategispel utan att kräva långa träningstider har visat sig vara en utmaning. Denna uppsats utforskar hur ett modellbaserat tillvägagångssätt, baserat på reglerna för spelet, kan användas för att skapa ett bra presterande program.

The cutting dates of 10 wood timbers (girders and corner rafters) of Sinmu-mun, the north gate of Kyungbok Palace in Seoul, were determined by the dendrochronological method. Tree-ring chronologies of unknown dates derived from the timbers were crossdated using the graphic comparison method against the dated master chronologies derived from living trees. The living trees for the masters used for this study were Pinus densiflora Sieb. et Zucc. (Japanese red pine), a major timber species for Korean traditional buildings. By comparing the Sinmu-mun samples with the masters from the western Sorak Mountains in central-eastern Korea, the Sinmu-mun samples yielded the cutting dates A.D. 1868, 1869, and 1870/1871. Surprisingly, these dates are 3 to 6 year later than the known date (A.D. 1865) of the Sinmu-mun reconstruction, which was recorded in a historical document ‘Ilsungrok’, the King’s official diary. Since the time that the Sinmu-mun construction date had been questioned, another record was found in the 1872 April issue of Ilsungrok, indicating the rebuilding of Sinmu-mun in the 1870s. Both pieces of evidence, from tree-ring dates and historic records, prove that the rebuilding of Sinmu-mun started after the Fall of 1870, but not later than April 1872. The results prove that tree-ring dating is a precise dating method and it can be applied to archaeological studies on Korean structures.

The game of Go is very ancient, with more than 4000 years of history and it is still popular nowadays, representing a big challenge to the Articial Intelligence. Despite its simple rules, the techniques which obtained success in other games like chess and draughts cannot handle satisfactorily the complex patterns and behaviours that emerge during a match of Go. The present work implements the SDM-Go, a competitive agent for Go that seeks to reduce the usage of supervision in the search process for the best move. The SDMGo utilizes the sparse distributed memory model as an additional resource to the Monte- Carlo tree search, which is used by many of the best automatic Go players nowadays. Based upon the open-source player Fuego, the use of the sparse distributed by SDM-Go has the purpose of being an alternative to the strong supervised process used by Fuego. The Monte-Carlo tree search executed by agent Fuego uses a set of heuristics codied by human professionals to guide the simulations and also to evaluate new nodes found in the tree. In a dierent way, SDM-Go implements a non-supervised and domain independent approach, where the history of the values of board states previously visited during the search are used to evaluate new boards (nodes of the search tree). In this way, SDM-Go reduces the supervision of Fuego, substituting its heuristics by the sparse distributed memory, which works as a repository for the information from the history of visited board states. Thus, the contributions of SDM-Go consist of: (1) the utilization of a sparse distributed memory to substitute the supervised approach of Fuego to evaluate new nodes found in the search tree; (2) the implementation of a board state representation based on bit vectors, in order to not compromise the performance of the system due to the boards stored in the memory; (3) the extension of the usage of the Monte-Carlo simulation results to update the values of the board states stored in the memory. Distinctly from many other existing agents, the use of the sparse distributed memory represents an approach independent of domain. The results obtained in tournaments against the well known open-source agent Fuego show that SDM-Go can perform successfully the task of providing a non-supervised and independent of domain approach to evaluate new nodes found in the search tree. Despite the longer runtime required by the use of the sparse distributed memory, the core of the agent performance, SDM-Go can keep a competitive level of play, especially at the 9X9 board.
Com mais de 4000 anos de história, o jogo de Go é atualmente um dos mais populares jogos de tabuleiro e representa um grande desao para a Inteligência Articial. Apesar de suas regras simples, as técnicas que anteriormente obtiveram sucesso em outros jogos como xadrez e damas não conseguem lidar satisfatoriamente com os padrões e comportamentos complexos que emergem durante uma partida de Go. O presente trabalho implementa o SDM-Go, um agente jogador de Go competitivo que procura reduzir a utilização de supervisão no processo de busca pelo melhor movimento. O SDM-Go emprega o modelo de memória esparsamente distribuída como um recurso adicional à busca em árvore Monte- Carlo utilizada por muitos dos melhores agentes automáticos atuais. Baseado no jogador código-aberto Fuego o uso da memória esparsamente distribuída pelo SDM-Go tem como objetivo ser uma alternativa ao processo fortemente supervisionado utilizado por aquele agente. A busca em árvore Monte-Carlo executada pelo jogador Fuego utiliza um conjunto de heurísticas codicadas por prossionais humanos para guiar as simulações e também avaliar novos nós encontrados na árvore. De maneira distinta, o SDM-Go implementa uma abordagem não supervisionada e independente de domínio, onde o histórico dos valores dos estados de tabuleiros previamente visitados durante a busca são utilizados para avaliar novos estados de tabuleiro (nós da árvore de busca). Desta maneira, o SDM-Go reduz a supervisão do agente Fuego, substituindo as heurísticas deste pela memória esparsamente distribuída que funciona como repositório das informações do histórico de estados de tabuleiro visitados. Assim, as contribuições do SDM-Go consistem em: (1) a utilização de uma memória esparsamente distribuída para substituir a abordagem supervisionada do Fuego para avaliar previamente novos nós encontrados na árvore; (2) a implementação de uma representação de tabuleiro baseada em vetores de bits, para não comprometer o desempenho do sistema em função dos tabuleiros armazenados na memória; (3) a extensão da utilização dos resultados das simulações Monte-Carlo para atualizar os valores dos tabuleiros armazenados na memória. Diferentemente de muitos outros agentes atuais, o uso da memória esparsamente distribuída representa uma abordagem independente de domínio. Os resultados obtidos em torneios contra o conhecido agente código-aberto Fuego mostram que o SDM-Go consegue desempenhar com sucesso a tarefa de prover uma abordagem independente de domínio e não supervisionada para avaliar previamente novos nós encontrados na árvore de busca. Apesar do maior tempo de processamento requerido pela utilização da memória esparsamente distribuída, peça central para o desempenho do agente, o SDM-Go consegue manter um nível de jogo competitivo, principalmente no tabuleiro 9X9.
Mestre em Ciência da Computação

This thesis presents a new technique for mapping combinational circuits to Fine-Grain Cellular-Architecture FPGAs. We represent the netlist as the binary tree with decision variables associated with each node of the tree. The functionality of the tree nodes is chosen based on the target FPGA architecture. The proposed tree restructuring algorithms preserve local connectivity and allow direct mapping of the trees to the cellular array, thus eliminating the traditional routing phase. Also, predictability of the signal delays is a very important advantage of the developed approach. The developed bus-assignment algorithm efficiently utilizes the medium distance routing resources (buses). The method is general and can be used for any Fine Grain CA-type FPGA. To demonstrate our techniques, ATMEL 6000 series FPGA was used as a target architecture. The area and delay comparison between our methods and commercial tools is presented using a set of MCNC benchmarks. Final layouts of the implemented designs are included. Results show that the proposed techniques outperform the available commercial tools for ATMEL 6000 FPGAs, both in area and delay optimization.

The new family of Field Programmable Gate Arrays, CLI 6000 from Concurrent Logic Inc realizes truly Cellular Logic. It has been mainly designed for the realization of data path architectures. However, the realizable logic functions provided by its macrocells and their limited connectivity call also for new general-purpose logic synthesis methods. The basic cell of CLi 6000 can be programmed to realize a two-input multiplexer ( A*B + C*B ), an AND/EXOR cell ( A*B Ea C ), or the basic 2-input AND, OR and EXOR gate. This suggests to using these cells for tree-like expansions. These "cellular logic" devices require regular connection patterns in the netlists resulting from logic synthesis. This thesis presents a synthesis tree searching program PROMPT, which generates AND/EXOR tree circuits from given Boolean functions. Such circuits have the property that the gate structures are AND/EXOR ( A *B EB C ), AND and EXOR which could be realized by the CLI6000 cells. Also, the connection. way in the circuit is that usually the output of one level gate is the input of the next level gate of the tree. This matches ideally to the architecture of the CLI6000 bussing network where the macrocells have only connections to their neighboring cells. PROMPT is based on the Davio expansions ( an equivalent of the Shannon expansions for the EXOR gates ) as its Boolean decomposition methods. The program includes three versions: exact version, heuristic version and fixed-variable version. The exact version of PROMPT generates the Permuted Reed-Muller Tree circuit which has the minimum number of gates. Such tree circuit is obtained by searching through all possible combinations of the expansion variable orders to get the one which needs the least number of gates. The heuristic version of PROMPT is designed to decrease the time complexity of the search algorithm when dealing with logic functions having many input variables. It generates a Permuted Reed-Muller Tree which may not have the minimum number of gates. However, the tree searching time in this version decreases tremendously compared to the time necessary in the exact version. The fix-variable version is developed to generate Reed-Muller Tree circuits. Such circuits will have the same expansion variables at the same tree level, so they can be easier routed after the placement to the CLI6000 chips. In short, the program PROMPT generates the PRM and RM tree circuits which are particularly well matched to both the realization of logic cell and connection structure of the CLI6000 device. Thus, the PRM and RM circuits can be easily placed and routed on the CLI6000 FPGAs.

Cette thèse aborde la problématique de la création d’intelligences artificielles (IA) contrôlant la prise de décision haut-niveau dans les jeux de stratégie. Ce type de jeux propose des environnements complexes nécessitant de manipuler de nombreuses ressources en faisant des choix d’actions dépendant d’objectifs à long terme. La conception de ces IA n’est pas simple car il s’agit de fournir une expérience pour le joueur qui soit divertissante et intéressante à jouer. Ainsi, le but n’est pas d’obtenir des comportements d’IA imbattables, mais plutôt de refléter différents traits de personnalités permettant au joueur d’être confronté à des adversaires diversifiés. Leur conception fait intervenir des game designers qui vont définir les différentes stratégies en fonction de l’expérience qu’ils souhaitent créer pour le joueur, et des développeurs qui programment et intègrent ces stratégies au jeu. La collaboration entre eux nécessite de nombreux échanges et itérations de développement pour obtenir un résultat qui correspond aux attentes des designers. L’objectif de cette thèse est de proposer une solution de modélisation de stratégies accessible aux game designers en vue d’améliorer et de simplifier la création de comportements stratégiques. Notre proposition prend la forme d’un moteur stratégique choisissant des objectifs à long terme et vient se placer au dessus d’un module tactique qui gère l’application concrète de ces objectifs. La solution proposée n’impose pas de méthode pour résoudre ces objectifs et laisse libre le fonctionnement du module tactique. Le moteur est couplé à un modèle de stratégie permettant à l’utilisateur d’exprimer des règles permettant au moteur de choisir les objectifs et de leur allouer des ressources. Ces règles permettent d’exprimer le choix d’objectifs en fonction du contexte, mais également d’en choisir plusieurs en parallèle et de leur donner des importances relatives afin d’influencer la répartition des ressources. Pour améliorer l’intelligibilité nous utilisons un modèle graphique inspiré des machines à états finis et des behavior trees. Les stratégies créées à l’aide de notre modèle sont ensuite exécutées par le moteur de stratégie pour produire des directives qui sont données au module tactique. Ces directives se présentent sous la forme d’objectifs stratégiques et de ressources qui leur sont allouées en fonction de leurs besoins et de l’importance relative qui leur a été donnée. Le module stratégique permet donc de rendre accessible la conception du niveau stratégique d’une IA contrôlant un adversaire dans un jeu de stratégie
This PhD thesis addresses the topic of creating artificial intelligence (AI) to control high-level decision-making in strategy games. This kind of game offers complex environments that require the manipulation of a large number of resources by choosing actions depending on long-term goals. This AI design is not simple because it is about providing to the player a playful and interesting experience. Hence, the aim is not to create unbeatable behaviors, but rather to display several personality traits allowing the player to face diverse opponents. Its creation involves game designers who are responsible of defining several strategies according to the experience they want to provide to the player, and game developers who implement those strategies to put them into the game. The collaboration between them requires many exchanges and development iterations to obtain a result corresponding to game designers’ expectations. The objective of this PhD thesis is to improve and simplify the creation of strategical behaviors by proposing a strategy model intelligible to game designers and that can be interfaced easily with developers’ work. For game designers, a strategy model has been created to allow them to express rules guiding the choice of goals and their allocated resources. These rules make it possible for game designers to express which goal to choose according to the context but also to choose several of them and give them relative importance in order to influence the resource distribution. To improve intelligibility we use a graphical model inspired from finite state machines and behavior trees. Our proposition also includes a strategy engine which executes the strategies created with the model. This execution produces directives that are represented by a list of selected strategical goals and the resources that have been allocated according to the importance and needs of each goal. These directives are intended for a tactical module in charge of their application. The developers are then responsible for the implementation of this tactical module. Our solution enables game designers to directly design the strategical level of an AI and therefore facilitates their cooperation with game developers and simplifies the entire creation process of the AI

The aim of this thesis has been to explore how messages in Twitch chats are affected by the gender of the streamer and the type of game that they are playing. Using a quantitative method, messages from twelve different streamers, male and female were downloaded and categorised depending on their content. The analysis used theories on game categorisation in order to understand the complexity of the games that the broadcasters were playing, and in combination with this knowledge and the results of the data collection conclusions could be drawn between the complexity of a game and the amount of comments regarding gameplay or appearance in female streams.The analysis used feminist theories in order to understand the underlying reasons for the observed exclusion of women in both the gaming world in general, and in the Twitch streams. This analysis showed tendencies in the Twitch audience to adapt a male gaze as they were more prone to comment on passive aspects of the female streamers. The thesis concluded that the "true gamer" culture seems to be affecting female streamer son Twitch in several ways, and that the exclusion of female streamers takes different forms depending on the game they are playing.

Thesis (M.S. in Operations Analysis)--Naval Postgraduate School, June 2003.
Thesis advisor(s): Thomas W. Lucas, Samuel E. Buttrey. Includes bibliographical references (p. 73). Also available online.

This thesis describes the development of methods for automatically obtaining coarse gaze direction estimates for pedestrians in surveillance video. Gaze direction estimates are beneficial in the context of surveillance as an indicator of an individual's intentions and their interest in their surroundings and other people. The overall task is broken down into two problems. The first is that of tracking large numbers of pedestrians in low resolution video, which is required to identify the head regions within video frames. The second problem is to process the extracted head regions and estimate the direction in which the person is facing as a coarse estimate of their gaze direction. The first approach for head tracking combines image measurements from HOG head detections and KLT corner tracking using a Kalman filter, and can track the heads of many pedestrians simultaneously to output head regions with pixel-level accuracy. The second approach uses Markov-Chain Monte-Carlo Data Association (MCMCDA) within a temporal sliding window to provide similarly accurate head regions, but with improved speed and robustness. The improved system accurately tracks the heads of twenty pedestrians in 1920x1080 video in real-time and can track through total occlusions for short time periods. The approaches for gaze direction estimation all make use of randomised decision tree classifiers. The first develops classifiers for low resolution head images that are invariant to hair and skin colours using branch decisions based on abstract labels rather than direct image measurements. The second approach addresses higher resolution images using HOG descriptors and novel Colour Triplet Comparison (CTC) based branches. The final approach infers custom appearance models for individual scenes using weakly supervised learning over large datasets of approximately 500,000 images. A Conditional Random Field (CRF) models interactions between appearance information and walking directions to estimate gaze directions for head image sequences.

Kort beskrivelse av prosjektet: Oppgaven har vært å tegne et samlingshus for et antall omgivende videregående skoler. Ved å sentralisere ulike funksjoner som bibliotek, sosiale arealer og matsal fungerer huset som et felles læresenter for elevene.  I en tidlig fase fant jeg ut at jeg ville fokusere på å gjøre huset tilgjengelig for offentligheten slik at det ikke blir stående tomt utenom skoletidene. Dette ville jeg uttrykke både i form og i tolking av programmet. Så hvordan viser man at en bygning er åpen for alle i form? Jeg valgte å ta inspirasjon fra det mest offentlige jeg kjenner til, nemlig det offentlige rommet. Det jeg anser som spesielt med dette er at rommet bare fortsetter og fortsetter og består ikke av stengte vegger som man ofte finner i konvensjonelle bygninger. Ved å la gulvet være sammenhengende gjennom hele huset oppnås et interiør som jeg tror kan oppfattes som et offentlig rom. Jeg forsterket denne opplevelsen ved å koble gulvet til fjellet som grenser til tomten og huset ble dermed en offentlig vei per definisjon.  Hvordan fungerer interiøret? En utfordring har vært å definere rom når konseptet er at det skal være helt åpent. Derfor har jeg valgt ut fire parametere for å definere funksjonene: Møbler, et flatt gulv, dobbel takhøyde og til slutt det stengte rommet (se diagram, romdefineringer, s. 5). Hvilken rolle spiller programmet i konseptet? Flere av funksjonene i programmet var hva man ofte tenker på som soner kun for ansatte, som for eksempel administrasjon, rengjøring og kjøkken Jeg valgte å utfordre denne ideen og gjøre det slik at disse funksjonene ble tilgjengelige for folk flest. Gjennom et internt bookingsystem kan selv de som ikke er elever anvende huset og dets muligheter. For eksempel kan man booke, utenom skoletider, kjøkkenet for å lage mat i større volumer, black boxen for å holde en forestilling, et singelrom som kontor, et konferanserom for lesesirkelen på søndager, det multifunksjonelle torget for en bryllupsfest, osv. Hva oppnår man ved å koble huset med Enskede Gård? I planprogrammet står det: ”Stockholms ytterstad har till stor del vuxit fram etappvis och resulterat i stadsdelar som sinsemellan skiljer sig åt, men som i sig själva har ett homogent bostadsbestånd med likartade typologier och prisnivåer. (...) Slakthusområdet, som är tänkt att omvandlas under drygt ett decennium, är en del av denna utveckling. (...) En mer differentierad prisbild och lägre hyresnivåer finns att hitta i det befintliga beståndet, vilket gäller både bostäder och lokaler. Därför spelar programmets ambition att koppla samman Slakthusområdet med omkringliggande stadsdelar en viktig roll i att minska den geografiska segregationen mellan de olika områdena och skapa en funktionell helhet av dagens separata delar”. Utover dette ligger Lindeparken gymnasiesärskola i området og kan med grepet ta del av læresenterets tilbud.  Materialer? Ved å bruke materialer som ofte forekommer i et byrom kan følelsen av et offentlig og åpent rom forsterkes. I huset forekommer derfor stein, asfalt, betong, glass og stål. I tillegg har jeg fokusert på å la grensen mellom ute og inne defineres kun med glass samtidig som den dras inn ved den offentlige veien. Utover dette har jeg latt de rommene som er stengte få en varmere karakter, slik at de blir som små hus man går inn i, fra gaten.
Project description: Three planned high schools in the area were to have a common building with different social functions, such as a library, social areas and a dining room.  Early on in the process, I figured that the program had many functions that potensially could be interesting for the general public, such as a library, social areas and an exhibition. This led to the concept of opening up the building and making it accessible to everyone, creating potensial for an active building, even after school hours.  So how can openness be communicated in form? My concept was to extend the public space and let it continue uninterrupted through the whole interior., in short: a ramp. By connecting the ramp to Enskede, lying ten metres above the plot, the building became a public street. Stockholm municipality had an ambition for opening up Slakthusomradet, and the building became a part of this project. The program in the assignment excisted of many zones, usually associated with staff areas. I wanted to change that conception and make these rooms open to the public. These rooms included for example administration, cleaning and kitchen.  The materials used were all taken from typical public spaces, such as concrete, stone, asphalt, steel and glass. Furthermore, the character of the materials were to change when you moved from “the street” and into the closed spaces. Here you would find wooden walls and furniture, as well as curtains of cotton.

In the domain of strategic game playing, the use of statistical techniques such as the Upper Confidence for Trees (UCT) algorithm, has become the norm as they offer many benefits over classical algorithms. These benefits include requiring no game-specific strategic knowledge and time-scalable performance. UCT does not incorporate any strategic information specific to the game considered, but instead uses repeated sampling to effectively brute-force search through the game tree or search space. The lack of game-specific knowledge in UCT is thus both a benefit but also a strategic disadvantage. Pattern recognition techniques, specifically Neural Networks (NN), were identified as a means of addressing the lack of game-specific knowledge in UCT. Through a novel hybridisation technique which combines UCT and trained NNs for pruning, the UCTNN algorithm was derived. The NN component of UCT-NN was trained using a UCT self-play scheme to generate game-specific knowledge without the need to construct and manage game databases for training purposes. The UCT-NN algorithm is outlined for pruning in the game of Go-Moku as a candidate case-study for this research. The UCT-NN algorithm contained three major parameters which emerged from the UCT algorithm, the use of NNs and the pruning schemes considered. Suitable methods for finding candidate values for these three parameters were outlined and applied to the game of Go-Moku on a 5 by 5 board. An empirical investigation of the playing performance of UCT-NN was conducted in comparison to UCT through three benchmarks. The benchmarks comprise a common randomly moving opponent, a common UCTmax player which is given a large amount of playing time, and a pair-wise tournament between UCT-NN and UCT. The results of the performance evaluation for 5 by 5 Go-Moku were promising, which prompted an evaluation of a larger 9 by 9 Go-Moku board. The results of both evaluations indicate that the time allocated to the UCT-NN algorithm directly affects its performance when compared to UCT. The UCT-NN algorithm generally performs better than UCT in games with very limited time-constraints in all benchmarks considered except when playing against a randomly moving player in 9 by 9 Go-Moku. In real-time and near-real-time Go-Moku games, UCT-NN provides statistically significant improvements compared to UCT. The findings of this research contribute to the realisation of applying game-specific knowledge to the UCT algorithm.

Cost automata are traditional finite state automata enriched with a finite set of counters that can be manipulated on each transition. Based on the evolution of counter values, a cost automaton defines a function from the set of structures under consideration to the natural numbers extended with infinity, modulo a boundedness relation that ignores exact values but preserves boundedness properties. Historically, variants of cost automata have been used to solve problems in language theory such as the star height problem. They also have a rich theory in their own right as part of the theory of regular cost functions, which was introduced by Colcombet as an extension to the theory of regular languages. It subsumes the classical theory since a language can be associated with the function that maps every structure in the language to 0 and everything else to infinity; it is a strict extension since cost functions can count some behaviour within the input. Regular cost functions have been previously studied over finite words and trees. This thesis extends the theory to infinite trees, where classical parity automata are enriched with a finite set of counters. Weak cost automata, which have priorities {0,1} or {1,2} and an additional restriction on the structure of the transition function, are shown to be equivalent to a weak cost monadic logic. A new notion of quasi-weak cost automata is also studied and shown to arise naturally in this cost setting. Moreover, a decision procedure is given to determine whether or not functions definable using weak or quasi-weak cost automata are equivalent up to the boundedness relation, which also proves the decidability of the weak cost monadic logic over infinite trees. The semantics of these cost automata over infinite trees are defined in terms of cost-parity games which are two-player infinite games where one player seeks to minimize the counter values and satisfy the parity condition, and the other player seeks to maximize the counter values or sabotage the parity condition. The main contributions and key technical results involve proving that certain cost-parity games admit positional or finite-memory strategies. These results also help settle the decidability of some special cases of long-standing open problems in the classical theory. In particular, it is shown that it is decidable whether a regular language of infinite trees is recognizable using a nondeterministic co-Büchi automaton. Likewise, given a Büchi or co-Büchi automaton as input, it is decidable whether or not there is a weak automaton recognizing the same language.