Bodies That Answer in Milliseconds: A Historical Development Review of Real-Time Physics Simulation in Game Engines

Abstract

This article offers a historical development review of real-time physics simulation, the engineering discipline whose development from the elasticity papers of the late 1980s through the impulse solvers, the particle methods, and the position-based dynamics of the present made falling, colliding, and cloth the default behaviors of the interactive world. The review reconstructs the development across six phases: the graphicists' inheritance, Demetri Terzopoulos's elastically deformable models and the physically based modeling curriculum, whose simulation enterprise computer graphics adopted from mechanics; the constraint problem, David Baraff's analytical non-penetration methods, whose solved contact forces ended the era of exploding stacks; the millisecond budget, Philip Hubbard's sphere-approximated collision detection and Brian Mirtich's impulse-based dynamics, whose time-critical methods translated the physics into the game loop; the game engine's translation, Thomas Jakobsen's Verlet particle character physics for Hitman and the course canon that taught the industry, whose fixes-by-projection valued stability over accuracy; the commercial maturation, Erin Catto's sequential impulses and the Box2D lineage, whose solved solver became the standard of the two-dimensional world and the textbooks of David Eberly and Ian Millington codified; and the position-based school, Matthias Muller's and Miles Macklin's methods and Jan Bender's survey, whose provable stability conquered soft bodies, fluids, and the modern engine. Three synthetic claims are advanced. First, game physics is the engineering of compromises, and its history is the history of deciding which physical truth to sacrifice. Second, the Verlet particle heresy of 2001 proved more durable than the accuracy it rejected. Third, the position-based settlement unified what the discipline had split. An agenda is proposed spanning differentiable simulation, learned dynamics, and the certification of believable physics.