Content uploaded by Antonio J. Fernández-Leiva
Author content
All content in this area was uploaded by Antonio J. Fernández-Leiva on Mar 02, 2015
Content may be subject to copyright.
A preview of the PDF is not available
Procedural Content Generation for Real-Time Strategy Games
Abstract and Figures
Videogames are one of the most important and profitable sectors in the industry of entertainment. Nowadays, the creation of a videogame is often a large-scale endeavor and bears many similarities with, e.g., movie production. On the central tasks in the development of a videogame is content generation, namely the definition of maps, terrains, non-player characters (NPCs) and other graphical, musical and AI-related components of the game. Such generation is costly due to its complexity, the great amount of work required and the need of specialized manpower. Hence the relevance of optimizing the process and alleviating costs. In this sense, procedural content generation (PCG) comes in handy as a means of reducing costs by using algorithmic techniques to automatically generate some game contents. PCG also provides advantages in terms of player experience since the contents generated are typically not fixed but can vary in different playing sessions, and can even adapt to the player herself. For this purpose, the underlying algorithmic technique used for PCG must be also flexible and adaptable. This is the case of computational intelligence in general and evolutionary algorithms in particular. In this work we shall provide an overview of the use of evolutionary intelligence for PCG, with special emphasis on its use within the context of real-time strategy games. We shall show how these techniques can address both playability and aesthetics, as well as improving the game AI.
Figures - uploaded by Antonio J. Fernández-Leiva
Author content
All figure content in this area was uploaded by Antonio J. Fernández-Leiva
Content may be subject to copyright.
... Previous research into PLG has explored its applicability to many different game genres. These include platform (Mourato, dos Santos, and Birra 2011), racing (Cardamone, Loiacono, and Lanzi 2011), role-playing (Valtchanov and Brown 2012), arcade (Cook and Colton 2011), stealth (Xu, Tremblay, and Verbrugge 2014), roguelike (Stammer et al. 2015) and real-time strategy (Lara-Cabrera et al. 2015). Several papers have also explored the use of PLG for physicsbased puzzle games, most notably for the Cut the Rope Copyright c 2016, Association for the Advancement of Artificial Intelligence (www.aaai.org). ...
This paper presents a procedural generation algorithm for levels in physics-based puzzle games similar to Angry Birds. The proposed algorithm creates levels consisting of various self-contained structures placed throughout a 2D area. Each structure can be placed either on the ground or atop floating platforms within the available level space. These structures are created using a variety of different block types and do not require predefined substructures or composite elements. Target object locations are determined based on a combination of factors, including structural protection, occupancy estimation and overall dispersion. Experiments were performed in order to determine the ideal input parameters for generating desirable levels. The expressivity of the generator was also evaluated and the results show that the proposed method can generate a wide variety of interesting levels.
... The most common types of "game content" that are generated are usually levels or sub-sections of levels, referred to as procedural level generation (PLG). PLG has been previously implemented in many different game types, including real-time strategy [5], [6], platform [7], racing [8], arcade [9], roleplaying [10], stealth [11] and rogue-like [12]. The General Video Game AI Competition has also worked on attempting to procedurally generate levels for multiple general games [13], [14], although the results so far are somewhat mixed. ...
This paper presents an overview of the second AIBIRDS level generation competition, held jointly at the 2017 IEEE Conference on Computational Intelligence and Games, and the 26th International Joint Conference on Artificial Intelligence. This competition tasked entrants with developing a level generator for the physics-based puzzle game Angry Birds. Submitted generators were required to deal with many physical reasoning constraints caused by the realistic nature of the game's environment, in addition to ensuring that the created levels were fun, challenging and solvable. This year's competition was a significant improvement over the previous year, with a greater number of participants and more advanced generators. Within this paper we describe the framework, rules, submitted generators and results for this competition. We also provide some background information on related research and other video game AI competitions, as well as discussing what can be learned from this year's competition. There are several game and real-world applications for this type of research, and we provide some examples of the types of levels we would like future competition entries to generate.
... Some examples of this kind of works are developing new gameplay mechanics capable of trigger new emotions on the players, and devising algorithms that adapt the game to the perceived emotional status of the player. Regarding this category, there is an interesting symbiosis between affective games and procedural content generation research communities [12], as the former could use the techniques developed by the latter as a way to automatically create new content for the videogame that is suitable for the emotional state of the player. ...
A taxonomy and state of the art revision on affective games
R Lara-Cabrera, D Camacho
Future Generation Computer Systems
online 1 January 2018
https://doi.org/10.1016/j.future.2017.12.056
... Some examples of this kind of works are developing new gameplay mechanics capable of trigger new emotions on the players, and devising algorithms that adapt the game to the perceived emotional status of the player. Regarding this category, there is an interesting symbiosis between affective games and procedural content generation research communities [12], as the former could use the techniques developed by the latter as a way to automatically create new content for the videogame that is suitable for the emotional state of the player. ...
Affective games are a sub-field of affective computing that tries to study how to design videogames that
are able to react to the emotions expressed by the player, as well as provoking desired emotions to them. To achieve those goals it is necessary to research on how to measure and detect human emotions using a computer, and how to adapt videogames to the perceived emotions to finally provoke them to the players. This work presents a taxonomy for research on affective games centring on the aforementioned issues. Here we devise as well a revision of the most relevant published works known to the authors on this area. Finally, we analyse and discuss which important research problem are yet open and might be tackled by future investigations in the area of affective games.
The development of artificial intelligence (AI) techniques that can assist with the creation and analysis of digital content is a broad and challenging task for researchers. This topic has been most prevalent in the field of game AI research, where games are used as a testbed for solving more complex real-world problems. One of the major issues with prior AI-assisted content creation methods for games has been a lack of direct comparability to real-world environments, particularly those with realistic physical properties to consider. Creating content for such environments typically requires physics-based reasoning, which imposes many additional complications and restrictions that must be considered. Addressing and developing methods that can deal with these physical constraints, even if they are only within simulated game environments, is an important and challenging task for AI techniques that intend to be used in real-world situations. The research presented in this thesis describes several approaches to creating and analysing levels for the physics-based puzzle game Angry Birds, which features a realistic 2D environment. This research was multidisciplinary in nature and covers a wide variety of different AI fields, leading to this thesis being presented as a compilation of published work. The central part of this thesis consists of procedurally generating levels for physics-based games similar to those in Angry Birds. This predominantly involves creating and placing stable structures made up of many smaller blocks, as well as other level elements. Multiple approaches are presented, including both fully autonomous and human-AI collaborative methodologies. In addition, several analyses of Angry Birds levels were carried out using current state-of-the-art agents. A hyper-agent was developed that uses machine learning to estimate the performance of each agent in a portfolio for an unknown level, allowing it to select the one most likely to succeed. Agent performance on levels that contain deceptive or creative properties was also investigated, allowing determination of the current strengths and weaknesses of different AI techniques. The observed variability in performance across levels for different AI techniques led to the development of an adaptive level generation system, allowing for the dynamic creation of increasingly challenging levels over time based on agent performance analysis. An additional study also investigated the theoretical complexity of Angry Birds levels from a computational perspective. While this research is predominately applied to video games with physics-based simulated environments, the challenges and problems solved by the proposed methods also have significant real-world potential and applications.
Finding the global best strategy for an autonomous agent (bot) in a RTS game is a hard problem, mainly because the techniques applied to do this must deal with uncertainty and real-time planning in order to control the game agents. This work describes an approach applying a Genetic Programming (GP) algorithm to create the behavioural engine of bots able to play a simple RTS. Normally it is impossible to know in advance what kind of strategies will be the best in the most general case of this problem. So GP, which searches the general decision tree space, has been introduced and used successfully. However, it is not straightforward what fitness function would be the most convenient to guide the evolutionary process in order to reach the best solutions and also being less sensitive to the uncertainty present in the context of games. Thus, in this paper three different evaluation functions have been proposed, and a detailed analysis of their performance has been conducted. The paper also analyses several aspects of the obtained bots, in addition to their final performance on battles, such as the evolution of the decision trees (behavioural models) themselves, or the influence on the results of noise or uncertainty. The results show that a victory-based fitness, which prioritises the number of victories, contributes to generate better bots, on average, than other functions based on other numerical aspects of the battles, such as the number of resources gathered, or the number of units generated.
Nowadays game engines are imperative for building 3D applications and games. This is for the reason that the engines appreciably reduce resources for employing obligatory but intricate utilities. This paper elucidates about a game engine, popular games developed by these engines and its foremost elements. It portrays a number of special kinds of contemporary game developed by engines in the way of their aspects, procedure and deliberates their stipulations with comparison.
This work presents a procedural content generation system that uses an evolutionary algorithm in order to generate interesting maps for a real-time strategy game, called Planet Wars. Interestingness is here captured by the dynamism of games (i.e., the extent to which they are action-packed). We consider two different approaches to measure the dynamism of the games resulting from these generated maps, one based on fluctuations in the resources controlled by either player and another one based on their confrontations. Both approaches rely on conducting several games on the map under scrutiny using top artificial intelligence (AI) bots for the game. Statistic gathered during these games are then transferred to a fuzzy system that determines the map’s level of dynamism. We use an evolutionary algorithm featuring self-adaptation of mutation parameters and variable-length chromosomes (which means maps of different sizes) to produce increasingly dynamic maps.
This paper explores the use of Hall-of-Fame (HoF) in the application of competitive coevolution for finding winning strategies in RobotWars, a two-player real time strategy (RTS) game developed in the University of Malaga for research purposes. The main goal is testing different approaches in order to implement the concept of HoF as part of the self learning mechanism in competitive coevolutionary algorithms. Five approaches were designed and tested, the difference between them being based on the implementation of HoF as a long or short-term memory mechanism. Specifically they differ on the police followed to keep the members in the champions' memory during an updating process which deletes the weakest individuals, in order to consider only the robust members in the evaluation phase. It is shown how strategies based on periodical update of the HoF set on the basis of quality and diversity provide globally better results.
This paper describes the application of competitive coevolution as a mechanism of self learning in a two-player real time strategy (RTS) game. The paper presents this (war) RTS game, developed by the authors as an open-source tool, and describes its (built-in) coevolutionary engine developed to find winning strategies. This engine applies a competitive coevolutionary algorithm that uses the concept of Hall-of- Fame to establish a long-term memory that is employed in the evaluation process. An empirical analysis of the performance of two different versions of this coevolutionary algorithm is conducted in the context of the RTS game. Moreover, the paper also shows, by an example, the potential of this coevolutionary engine as a prediction tool by inferring the initial conditions (i.e. army configuration) under which a battle has been executed when we know the final result.
The Google Artificial Intelligence (AI) Challenge is an international contest the objective of which is to program the AI in a two-player real time strategy (RTS) game. This AI is an autonomous computer program that governs the actions that one of the two players executes during the game according to the state of play. The entries are evaluated via a competition mechanism consisting of two-player rounds where each entry is tested against others. This paper describes the use of competitive coevolutionary (CC) algorithms for the automatic generation of winning game strategies in Planet Wars, the RTS game associated with the 2010 contest. Three different versions of a prime algorithm have been tested. Their common nexus is not only the use of a Hall-of-Fame (HoF) to keep note of the winners of past coevolutions but also the employment of an archive of experienced players, termed the hall-of-celebrities (HoC), that puts pressure on the optimization process and guides the search to increase the strength of the solutions; their differences come from the periodical updating of the HoF on the basis of quality and diversity metrics. The goal is to optimize the AI by means of a self-learning process guided by coevolutionary search and competitive evaluation. An empirical study on the performance of a number of variants of the proposed algorithms is described and a statistical analysis of the results is conducted. In addition to the attainment of competitive bots we also conclude that the incorporation of the HoC inside the primary algorithm helps to reduce the effects of cycling caused by the use of HoF in CC algorithms.
We consider search-based procedural content generation in the context of Planet Wars, an RTS game. The objective of this work is to generate maps for the aforementioned game, that result in an interesting game-play. In order to characterize interestingness we focus on the properties of balance and dynamism. The former captures the fact that no player is overwhelmed by the opponent during the game, whereas the latter tries to model the fact that there is a lot of action during the game. To measure these properties on a given map, we conduct several games on them using top AI bots and collect statistics which are, in turn, used as inputs of a fuzzy rule base. This system is embedded within an evolutionary algorithm that features self-adaptation of mutation parameters as well as variable-length chromosomes (thus implying maps of different sizes). The experimentation focuses both on the optimization of balance and dynamism as stand-alone properties and in the analysis of the different tradeoffs attainable through them. To reach this goal a multi objective approach is used. We analyze both the usefulness of map-size self-adaptation in each scenario, as well as the properties of maps leading to different tradeoffs between dynamism and balance.
This paper presents a procedural content generation (PCG) method that is able to generate aesthetic maps for a real-time strategy game. The maps were characterized based on either their geometrical properties or their topological measures (obtained in this latter case from the sphere-of-influence graph induced by each map). Using these features, a distance function between maps can be defined. This function is used in turn to determine how close/far each map generated by the PCG method (a self-adaptive evolutionary algorithm) is to a collection of maps which were taken initially to be aesthetic or non-aesthetic. This correspondence guided a multi-objective evolutionary approach whereby maps close to aesthetic maps and far to non-aesthetic maps are sought. Self-organizing maps are used to ascertain whether the so-generated maps naturally cluster together with aesthetic maps, as well as to provide a qualitative assessment of the ability of each set of features to characterize the latter.
This paper attempts to give a high-level overview of the field of artificial and computational intelligence (AI/CI) in games, with particular reference to how the different core research areas within this field inform and interact with each other, both actually and potentially. We identify ten main research areas within this field: NPC behavior learning, search and planning, player modeling, games as AI benchmarks, procedural content generation, computational narrative, believable agents, AI-assisted game design, general game artificial intelligence and AI in commercial games. We view and analyze the areas from three key perspectives: 1) the dominant AI method(s) used under each area; 2) the relation of each area with respect to the end (human) user; and 3) the placement of each area within a human-computer (player-game) interaction perspective. In addition, for each of these areas we consider how it could inform or interact with each of the other areas; in those cases where we find that meaningful interaction either exists or is possible, we describe the character of that interaction and provide references to published studies, if any. We believe that this paper improves understanding of the current nature of the game AI/CI research field and the interdependences between its core areas by providing a unifying overview. We also believe that the discussion of potential interactions between research areas provides a pointer to many interesting future research projects and unexplored subfields.
The self-organized map, an architecture suggested for artificial
neural networks, is explained by presenting simulation experiments and
practical applications. The self-organizing map has the property of
effectively creating spatially organized internal representations of
various features of input signals and their abstractions. One result of
this is that the self-organization process can discover semantic
relationships in sentences. Brain maps, semantic maps, and early work on
competitive learning are reviewed. The self-organizing map algorithm (an
algorithm which order responses spatially) is reviewed, focusing on best
matching cell selection and adaptation of the weight vectors.
Suggestions for applying the self-organizing map algorithm,
demonstrations of the ordering process, and an example of hierarchical
clustering of data are presented. Fine tuning the map by learning vector
quantization is addressed. The use of self-organized maps in practical
speech recognition and a simulation experiment on semantic mapping are
discussed