Behavior Trees for Computer Games

Authoring the AI for non-player characters (NPCs) in modern video games is an increasingly complex task. Designers and programmers must collaborate to resolve a tension between believable agents with emergent behaviours and scripted story lines. Behaviour trees (BTs) have been proposed as an expressive mechanism that let designers create complex behaviours along the lines of the story they want to tell. However, BTs are still too complex for non-programmers. In this paper, we propose the use of plan traces to assist designers when building BTs. In order to make this approach feasible within state-of-the-art video game technology, we generate the planning domain through an extension of the component-based approach, a widely used technique for representing entities in commercial video games.

First presented as a tool to develop intelligent agents in computer games, Behavior Trees (BTs) are an alternative to Finite State Machines that later gained popularity in robotics, due to its modularity and how easy it makes to create intelligent reactive behaviors. Even with BTs being popular, the available tools to design it still lack features that allow validation without the end-application, whether it is a game or a robot. In this paper, a novel tool for BTs design is presented. This tool allows us to create BTs and also to evaluate then without any external software, making it easy to check all possible behaviors during the design stage. Also, the BTs created in the software can be directly used in BehaviorTree.CPP, a popular behavior tree library for robotics.

As a reactive and modular policy control architecture, Behavior Tree (BT) has been used in computer games and robotics for autonomous agents' task switching. However, constructing BTs manually for complex tasks requires expert domain-knowledge and is error-prone. As a solution, researchers have proposed to auto-construct BTs using evolutionary algorithms such as Genetic Programming (GP) and Grammatical Evolution (GE). Nevertheless, their effectiveness in practical situations is in doubt and there are different drawbacks in the application.

Behavior Trees are a promising approach to model the autonomous behaviour of robots in dynamic environments. Behavior Trees represent action selection decisions as a tree of decision nodes. The hierarchy of these decision nodes provides the planning of actions of the robot including its reactions on exceptions. Behavior Trees enable flexible planning and replanning of robot behavior while supporting better maintainable decision-making than traditional Finite State Machines. This paper presents an overview of lessons, which we have learned when applying Behavior Trees to various autonomous robots. We present these lessons as a sequence of steps that is meant to support robot software practitioners to develop their systems.

Miyake Yōichirō, an engineer who creates artificial intelligence (AI) for characters in computer games, asserts that current AI technology is rooted in Western philosophical traditions. He argues that when Western AI was created it was not generally acknowledged as intelligent in its own right. It is therefore subjected to verification tests, such as the Turing test, and only if it passes these does it qualify as intelligent. However, Miyake points out that the goal of an AI engineer is to humanise AI by giving it desires and worries. He claims that it is necessary to merge Eastern and Western thought in the design process in order to achieve a more human-like holistic AI. Consequently, the configuration of AI in computer games depends on the perception of the nature of human intelligence held by the game designer. This article examines Miyake’s philosophical approach in defining and producing AI, and how the AI of computer game characters is influenced by the software developer’s view of human beings.

Behavior Tree (BT) is a behavior modeling tool derived from computer game design. It has the advantages of modularity, hierarchy, reactivity, and readability, which can be used to model the behavior of agents and implement task-switching functionalities. Parallel task execution in behavior trees may encounter resource conflicts similar to those in concurrent programming. To solve the problem of resource conflicts in parallel executions of BT, the actions are usually set execution priority in advance. However, this method is not optimal for scheduling parallel tasks in many BT applications. In this paper, we propose to integrate several classic resource scheduling algorithms into a popular BT framework and evaluate it by multiple experiments. The BT resource scheduling mechanism expands the architecture of the classic behavior tree.

Increasingly, digital systems interfere into our daily lives. However, some tasks remain to be assigned to biological organisms. Seeing eye dogs have not yet been functionally replaced by assistive technology; digital pets are still less commonly the recipients of human affection than real pets. Both science and arts have experimented amply with integration of biological entities and technological systems, with the goal of complex task achievement. Cockroaches, neural cells, fish and slime molds were employed to control robots. Communities of real crickets were shown to interact with robots. Pigeons, bats and dogs were used as sensory and decision-making parts in technological systems. Crickets, hamsters and paramecia have driven the behavior of non-player characters in computer games. Through crowdsourcing, even humans serve as key components in greater digital systems.

Americans are increasingly concerned about video games, presumably due to the amount and graphicness of violence they contain. Social Cognitive Theory suggests that people are more likely to imitate characters they see as attractive or similar to self. To date, however, little research has examined attributes of violent characters in video games related to this issue. This content analysis examined 10 minutes of play from 60 of the most popular video games. Adapting the coding scheme from the National Television Violence Study (Smith et al., 1998; Wilson et al., 1997, 1998), various demographic and contextual features of violent characters and violent interactions were assessed. The results show that violent game characters have attributes that are likely to increase the extent to which some players perceive them as attractive and similar. Moreover, these violent characters engage in aggression that is presented as justified or graphic.

Previous research has shown that automatically combining low-level behaviors into a probabilistic finite state machine produces control software that crosses the reality gap satisfactorily. In this paper, we explore the possibility of adopting behavior trees as an architecture for the control software of robot swarms. We introduce Maple: an automatic design method that combines preexisting modules into behavior trees. To highlight the potential of this control architecture, we present robot experiments in which we compare Maple with Chocolate and EvoStick on two missions: foraging and aggregation. Chocolate and EvoStick are two previously published automatic design methods. Chocolate is a modular method that generates probabilistic finite state machines and EvoStick is a traditional evolutionary robotics method. The results of the experiments indicate that behavior trees are a viable and promising architecture to automatically generate control software for robot swarms.

Behavior Trees (BTs) were first conceived in the computer games industry as a tool to model agent behavior, but they received interest also in the robotics community as an alternative policy design to Finite State Machines (FSMs). The advantages of BTs over FSMs had been highlighted in many works, but there is no thorough practical comparison of the two designs. Such a comparison is particularly relevant in the robotic industry, where FSMs have been the state-of-the-art policy representation for robot control for many years. In this work we shed light on this matter by comparing how BTs and FSMs behave when controlling a robot in a mobile manipulation task. The comparison is made in terms of reactivity, modularity, readability, and design. We propose metrics for each of these properties, being aware that while some are tangible and objective, others are more subjective and implementation dependent. The practical comparison is performed in a simulation environment with validation on a real robot. We find that although the robot’s behavior during task solving is independent on the policy representation, maintaining a BT rather than an FSM becomes easier as the task increases in complexity.

This paper firstly seeks an answer to the question of whether Non Player Characters (NPCs) in Computer Games can also be viewed as game agents where reactivity, autonomy, being temporarily continuous and having goal-oriented behavior are taken as the features of being a game agent. Within this scope, we will try to assess whether naming NPCs as agents point to a desire that one day they would fulfill the requirements of being an agent or whether NPCs actually already fulfill these requirements. Secondly, the paper looks into the AI needs of video games. We present the AI methodologies that are either being used and under research for use in Game AI. The same methodologies are also likely to contribute to the increasing levels of autonomy and goal-directed behavior of NPCs and help them become more agent-like.

Typically, the creation of AI-based behaviors for Non-Playing Characters (NPC) in computer games is carried out by people with specialized skills in the broad areas of design and programming. This book chapter presents Second Mind (SM), a digital solution that makes it possible for novice users to successfully author behaviors. In SM, authors can use a graphical interface to define the behaviors of virtual characters in response to interactions with players. Authors can easily endow NPCs with behavior capabilities that make them act in different possible roles such as e.g. shopkeepers, museum hosts, etc. A series of user tests with human subjects to evaluate the behavior authoring process shows that Second Mind is easy to understand and simplifies the process of behavior production.

The article presents the results of a study on the possibilities of building software behavioral components based on large language models (LLMs) and their use in gaming systems. The subject of the research is the behavioral component of non-player characters (NPCs) in computer games, integrated with an LLM. The aim is to explore and analyze the features of applying LLMs to simplify NPC development. The study examines game AI types – reactive, state-driven, and learning-based. Interaction with NPCs through LLMs for chat communication is considered, and analogues of such characters’ behavior and dialogue are explored. The outcome is a designed and implemented API service for interaction with LLMs and an external gaming system with a corresponding software behavioral component that provides functionality for LLM-based interaction. Testing has shown that this system works with 54 unique contextual properties, achieves a maximum system prompt generation time of up to 60 ms, a maximum LLM response parsing and validation time of up to 30 ms, and supports models with at least 14 billion parameters. Keywords: API, LLM, NPC, XML, contextual action, contextual property, software behavioral component.

Studies on charactonym leave an underinvestigated field of moves or jutsu in Japanese. Characters in action comics, anime, and computer games are equipped with moves e.g. Goku with his iconic Kamehameha, Monkey D. Luffy with his Gomu Gomu no Pistol, and Spider Man with his Maximum Spider. We argue that these moves deserve their own umbrella in charactonym studies – we call it jutsunym. Though adopting a Japanese martial art term, jutsunym is not limited on Japanese media. Investigating a corpus of jutsu defining comic, anime, and game series through the lens of Smith’s Peircean names as signs and Rudnyckyj’s relevance of content and form which was later developed by Gerus-Tarnawecky’s theory in literary onomastics, we found that in naming a jutsu or a move, the authors tend to consider what we call as indexical alignment – an alignment between the names of the jutsu or the moves with narrative, mechanical, visual, and audial elements. Narrative elements align themselves with jutsu names through story and character, mechanical elements through connotative and denotative, visual elements through indicative and implicative, and audial elements through quoting and catchphrasing.

Over the last several decades, computer games started to have a significant impact on society. However, although a computer game is a type of software, the process to conceptualize, produce and deliver a game could involve unusual features. In software testing, for instance, studies demonstrated the hesitance of professionals to use automated testing techniques with games, due to the constant changes in requirements and design, and pointed out the need for creating testing tools that take into account the flexibility required for the game development process. Goal. This study aims to improve the current body of knowledge regarding software testing in game development and point out the existing particularities observed in software testing considering the development of a computer game. Method. A mixed-method approach based on a case study and an opinion survey was applied to collect quantitative and qualitative data from software professionals regarding the particularities of software testing in game development. Results. We analyzed over 70 messages posted on three well-established network of question-and-answer communities related to software engineering, software testing and game development and received answers of 38 professionals discussing differences between testing a computer game and a general software, and identified important aspects to be observed by practitioners in the process of planning, performing and reporting tests in this context. Conclusion. Considering computer games, software testing must focus not only on the common aspects of a general software, but also, track and investigate issues that could be related to game balance, game physics and entertainment related-aspects to guarantee the quality of computer games and a successful testing process.