"Why Halo Infinite's Bots Play More Like Humans | AI and Games #71" is a video published by AI and Games that delves into the development of AI-controlled bots in the Halo Infinite multiplayer modes. The video discusses how 343 Industries created bots that behave more like human players using a combination of traditional AI techniques and a unique approach to designing their behaviors.
Halo Infinite is the first entry in the series to offer a free-to-play multiplayer model, attracting a broader range of players with varying experiences. To cater to these diverse players, 343 Industries focused on onboarding and improving the overall experience for both novice and seasoned players. Sarah Stern, the Senior Multiplayer Designer for Halo Infinite, led a team of designers and programmers to redefine onboarding in Halo Infinite, including the development of bots and the Academy, a new tutorial system that utilizes the bots.
Contrary to expectations, the bots in Halo Infinite are not built using machine learning or deep neural networks but are instead developed using traditional game AI and behavior trees. The bots exhibit common behavioral traits expected of human players, such as engaging in combat, interacting with objectives, and traversing the gameplay space.
The bots' human-like behavior is achieved through three design pillars: individual skills, a utility AI system within the behavior trees, and a nuanced understanding of game modes. The individual skills of the bots, such as strafing, aiming, grenade usage, and melee combat, are designed to appear more human and fallible while also adjusting across multiple skill levels. The utility AI system allows the bots to react to the current state of the game and prioritize actions based on human play styles. Lastly, the bots have a more nuanced understanding of game modes, which factors into their decision-making.
To implement the human-like skills, the bot team developed levels of performance that reflect how human players actually play Halo multiplayer. These levels are transposed against the four difficulty levels of the bots: Recruit, Marine, ODST, and Spartan. Each skill is designed to evolve over time, reflecting the gradual improvement of human players as they spend more time with the game.
The utility AI system used by the bots analyzes the current game state and prioritizes specific behaviors in the behavior tree based on ambition values. These values are calculated by factoring in game state information and normalizing it to a value between 0 and 1. The higher the value, the more desirable the ambition is deemed to be. This system allows the bots to make decisions that align with the context of the Halo combat dance and balance between fighting enemy players and winning the match in specific game modes.
Additional considerations for the bots include preventing thrashing, where the AI constantly switches between two actions, and sharing information among bots on higher difficulty levels. The Lua programming language is used for the utility calculations, which proved to be fast enough for the bots to process information in a similar time window as human players.
The development of human-like bots in Halo Infinite not only provides a solution for uneven lobbies but also helps players develop and improve their skills in the game. By replicating core skills that human players develop and incorporating a utility AI system that reacts to the game state, 343 Industries has created AI-controlled Spartans that offer a more engaging and authentic multiplayer experience.
