Flocking with Swift by Paul Hudson

The Evolution of Game AI

• 💡 Early video game AI relied on complex if-statements to respond to player actions, which became predictable once players learned the rules.
• ⚠️ Adding more complexity led to bugs and unintentional behaviors, like the "nuclear Gandhi" myth in Civilization, where a buffer overflow supposedly made a peaceful leader aggressive.
• 🎲 Developers found that replacing complex rules with random number generators (rolling a dice) made AI less predictable and more enjoyable for players, who would interpret meaning into the randomness.

Understanding Flocking Principles

• 🧠 The video introduces flocking, a simple AI technique used in computer graphics to simulate realistic group movements like birds or fish.
• 🔑 Flocking behavior is governed by three fundamental patterns: cohesion (staying near the flock's average position), separation (maintaining a small distance from others), and alignment (matching the average direction of the flock).
• ✨ These individually simple rules combine to create complex and beautiful swarm behaviors, first described by Craig Reynolds in 1987 as "Boids."

Swift Implementation Details

• 🛠️ Implementing Boids in Swift involves creating a Void class for individual objects (position, velocity, max speed) and a Flock class to manage a collection of voids.
• 🚀 The SIMD framework (Single Instruction, Multiple Data) is used for fast mathematical calculations on large amounts of data, crucial for performance with many voids.
• 🎨 SwiftUI is utilized for rendering, drawing individual voids as triangles and updating their positions and orientations regularly using a timeline view and Canvas.

Enhancing Flocking Dynamics

• ✅ The implementation includes helper methods like nearby to find neighboring voids efficiently and steering to smooth out movement adjustments.
• 👥 Team functionality allows voids to flock only with others of the same team, creating multiple distinct groups on screen.
• 🚧 Object avoidance is added, enabling voids to dynamically move away from a designated obstacle point, which can be controlled by user interaction.

The Power of Simple AI

• 💡 This approach demonstrates how simple, iterative logic can produce complex and beautiful emergent behaviors that mimic real-world phenomena.
• ⚡ The technique is easy to understand and extremely fast, even on older hardware, making it a powerful example of effective AI.
• 🌱 Future possibilities include adding multiple obstacles, extending to 3D space, or incorporating predators to further enrich the simulation.