Textbook in PDF format

Game AI Uncovered: Volume One kicks off a brand-new series of books that focus on the development of artificial intelligence in video games. This volume brings together the collected wisdom, ideas, tricks, and cutting-edge techniques from 20 of the top game AI professionals and researchers from around the world.
The techniques discussed in these pages cover the underlying development of a wide array of published titles, including Hood: Outlaws and Legends, The Escapists 2, Sackboy: A Big Adventure, Call of Duty: Strike Team, GTI+ Club, Split/Second, Sonic All Stars Racing Transformed, Luna Abyss, Medal of Honor Heroes I & II, Age of Empires IV, Watch Dogs, Battlefield 2042, Plants vs. Zombies: Battle for Neighborville, Dead Space, and more.
Contained within this volume are overviews and insight covering a host of different areas within game AI, including situational awareness, pathfinding, tethering, squad behaviours, coordination, auto-generating navigation link data, fluid movement, combining behaviour and animation systems, pedal control for cars, tactical positioning, level of detail, infinite axis utility systems, hierarchical state machines, bots for testing, reactive behaviour trees, and more.
For me, game artificial intelligence is by far the most interesting part of game development. That is not to disparage the great work that is done in graphics, physics, tools, audio, and all the other vital components required to make a game, but there is something special about bringing an agent to life. It is not only about the end result though, but also about how you get there. The smoke and mirrors that are used for a player to buy into the illusion and keep them there. Whether it is finding a believable path across an environment, navigating a busy street, communication between agents, or building an architecture that handles everything the design department can throw at it. It is always a journey of discovery. You start with ideas, and these evolve as the game grows.
Hood: Outlaws and Legends is a third-person, multiplayer action game based in mediaeval England. The primary game mode is a heist style mode with up to eight players split across two teams that try to steal treasure from an AI-controlled army. The AI army is led by the tyrannous Sheriff and consists of melee and ranged guards, with more than one hundred characters spread across the map. The large numbers of disparate AI characters, and the fact that players could be individually spread across the map, meant that the AI framework had to be robust and efficient to ensure that the game was as immersive as possible. Furthermore, the servers running Hood were chosen to be as cost effective as possible, which meant that the amount of memory and processor power available were very limited. Hood was developed in Unreal Engine 4 (UE4) and shipped on all major platforms at the time; however, all the AI logic would be run on low-cost AWS (Amazon Web Services) server instances. This chapter will cover some of the design choices and limitations that were overcome to deliver the desired player experience. Whilst these techniques were applied specifically to get the AI running efficiently in Hood, some of them should be applicable to other areas of gameplay development too.
Preface
The Changing Landscape of AI for Game Development
Implementing an AI Framework in Unreal for Medium- to Large-Scale Numbers of Disparate AI in Hood: Outlaws and Legends
Predicting When and Where to Intercept Players in Hood: Outlaws and Legends
Getting Caught Off Guard: Situational Awareness in The Escapists 2
Stitching It Together: The Enemies of Sackboy: A Big Adventure
Squad Behaviour Using a Task Stack Approach for Call of Duty: Strike Team
Reactive Behaviour Trees
Tailoring Racing AI to the Genre
Believable Routes: A Pathfinding Acceptability Metric
AI Under Lockdown: Various Approaches to NPC Tethering
Low-Cost, Mutable Path-Planning Spaces
Auto-Generating Navigation Link Data
Fluid Movement
Combining Behaviour and Animation Systems in Luna Abyss
How to Make Your AI Fireproof: 25 Years of AI Firefighting
Pedal Control for Cars: How to Maintain Speed, Change Speed, and Stop at a Point.
Bots for Testing
Infinite Axis Utility System and Hierarchical State Machines
Tactical Positioning through Environmental Analysis
Level of Detail: Saving Crucial CPU Time
Agent Coordination: Designing a Generalised Coordination Layer