Monte-Carlo Tree Search (MCTS) is an online planning algorithm that combines the ideas of best-first tree search and Monte-Carlo evaluation. Since MCTS is based on sampling, it does not require a transition function in explicit form, but only a generative model of the domain. Because it grows a highly selective search tree guided by its samples, it can handle huge search spaces with large branching factors. By using Monte-Carlo playouts, MCTS can take long-term rewards into account even with distant horizons. Combined with multi-armed bandit algorithms to trade off exploration and exploitation, MCTS has been shown to guarantee asymptotic convergence to the optimal policy, while providing approximations when stopped at any time.

Heuristic search has been very successful in abstract game domains such as Chess and Go. In video games, however, adoption has been slow due to the fact that state and move spaces are much larger, real-time constraints are harsher, and constraints on computational resources are tighter. In this paper we present a fast search method -- Alpha-Beta search for durative moves-- that can defeat commonly used AI scripts in RTS game combat scenarios of up to 8 vs. 8 units running on a single core in under 5ms per search episode. This performance is achieved by using standard search enhancements such as transposition tables and iterative deepening, and novel usage of combat AI scripts for sorting moves and state evaluation via playouts. We also present evidence that commonly used combat scripts are highly exploitable -- opening the door for a promising line of research on opponent combat modelling.

Planning-based techniques are a very powerful tool for automated story generation. However, as the number of possible actions increases, traditional planning techniques suffer from a combinatorial explosion due to large branching factors. In this work, we apply Monte Carlo Tree Search (MCTS) techniques to generate stories in domains with large numbers of possible actions (100). Our approach employs a Bayesian story evaluation method to guide the planning towards believable stories that reach a user defined goal. We generate stories in a novel domain with different type of story goals. Our approach shows an order of magnitude improvement in performance over traditional search techniques.

In order to automatically generate high-quality game levels, one needs to be able to automatically verify that the levels are playable. The simulation-based approach to playability testing uses an artificial agent to play through the level, but building such an agent is not always an easy task and such an agent is not always readily available. We discuss this prob- lem in the context of the physics-based puzzle game Cut the Rope, which features continuous time and state space, mak- ing several approaches such as exhaustive search and reactive agents inefficient. We show that a deliberative Prolog-based agent can be used to suggest all sensible moves at each state, which allows us to restrict the search space so that depth-first search for solutions become viable. This agent is successfully used to test playability in Ropossum, a level generator based on grammatical evolution. The method proposed in this paper is likely to be useful for a large variety of games with similar characteristics.

Game tree search in games with large branching factors is a notoriously hard problem. In this paper, we address this problem with a new sampling strategy for Monte Carlo Tree Search (MCTS) algorithms, called "Naive Sampling", based on a variant of the Multi-armed Bandit problem called the "Combinatorial Multi-armed Bandit" (CMAB) problem. We present a new MCTS algorithm based on Naive Sampling called NaiveMCTS, and evaluate it in the context of real-time strategy (RTS) games. Our results show that as the branching factor grows, NaiveMCTS performs significantly better than other algorithms.

From an AI point of view, Real-Time Strategy (RTS) games are hard because they have enormous state spaces, they are real-time and partially observable. In this paper, we explore an approach to deploy game-tree search in RTS games by using game state abstraction, and explore the effect of using different abstractions over the game state. Different abstractions capture different parts of the game state, and result in different branching factors when used for game-tree search algorithms. We evaluate the different representations using Monte Carlo Tree Search in the context of StarCraft.

From an AI point of view, Real-Time Strategy (RTS) games are hard because they have enormous state spaces, they are real-time and partially observable. In this paper, we present an approach to deploy game-tree search in RTS games by using game state abstraction. We propose a high-level abstract representation of the game state, that significantly reduces the branching factor when used for game-tree search algorithms. Using this high-level representation, we evaluate versions of alpha-beta search and of Monte Carlo Tree Search (MCTS).

Standard game tree search algorithms, such as minimax or Monte Carlo Tree Search, assume the existence of an accurate forward model that simulates the effects of actions in the game. Creating such model, however, is a challenge in itself.One cause of the complexity of the task is the gap in level of abstraction between the informal specification of the model and its implementation language. To overcome this issue, we propose a technique for the implementation of forward models that relies on the Answer Set Programming paradigm and on well-established knowledge representation techniques from defeasible reasoning and reasoning about actions and change. We evaluate our approach in the context of Real-Time Strategy games using a collection of StarCraft scenarios.

Game tree search algorithms, such as Monte Carlo Tree Search (MCTS), require access to a forward model (or "simulator") of the game at hand. However, in some games such forward model is not readily available. In this paper we address the problem of automatically learning forward models (more specifically, combats models) for two-player attrition games. We report experiments comparing several approaches to learn such combat model from replay data to models generated by hand. We use StarCraft, a Real-Time Strategy (RTS) game, as our application domain. Specifically, we use a large collection of already collected replays, and focus on learning a combat model for tactical combats.

IW(1) is a simple search algorithm that assumes that states can be characterized in terms of a set of boolean features or atoms. IW(1) consists of a standard breadth-first search with one variation: a newly generated state is pruned if it does not make a new atom true. Thus, while a breadth-first search runs in time that is exponential in the number of atoms, IW(1) runs in linear time. Variations of the algorithm have been shown to yield state-of-the-art results in classical planning and more recently in the Atari video games. In this paper, we use the algorithm for selecting actions in the games of the general video-game AI competition (GVG-AI) which, unlike classical planning problems and the Atari games, are stochastic. We evaluate a variation of the algorithm over 30 games under different time windows using the number of wins as the performance measure. We find that IW(1) does better than the sample MCTS and OLMCTS controllers for all time windows with the performance gap growing with the window size. The exception are the puzzle-like games where all the algorithms do poorly. For such problems, we show that much better results can be obtained with the IW(2) algorithm, which is like IW(1), except that states are pruned in the breadth-first search when they fail to make true a new pair of atoms.