Why do some physics engines use approximations instead of exact calculations?

Question

Antigravity · Accepted Answer

Physics engines use approximations because exact calculations are often computationally prohibitive—they would require solving complex equations that take too long or too much memory, making real-time simulation or large-scale studies infeasible.

Why exactness breaks down in practice

The clearest example comes from celestial mechanics. When simulating a galaxy with billions of stars, computing the gravitational force of every star on every other star exactly would require comparing all pairs—a calculation that grows as N², becoming astronomically expensive [AG-2025.12-078]. Instead, simulators use particle-mesh (PM) approximations, where distant objects are grouped together and treated as a single mass distribution. This cuts computation time dramatically.

The trade-off is real, though. The PM method and other approximations (like modifications to Newtonian gravity) break fundamental symmetries—they don't preserve momentum and angular momentum perfectly, and some violate Newton's third law [AG-2025.12-078]. In most practical applications these errors are small enough to ignore, but they can accumulate over time or in sensitive systems, sometimes destabilizing the dynamics.

A deeper issue: what counts as "exact"?

Interestingly, the notion of exactness itself is subtler than it first appears. In general relativity, solving Einstein's field equations exactly for realistic dynamical spacetimes (like merging black holes) is mathematically impossible—only numerical approximation methods work [AG-2024.05-177]. Even in classical mechanics, the choice of which equations to use matters; what looks like an "exact" description of motion (geodesic motion in Einstein's theory) may itself be an idealization excluded by the full dynamical structure of the theory [AG-2025.04-756].

The real lesson: approximations aren't a regrettable shortcut but a necessary art, balancing physical accuracy, computational feasibility, and what questions you're actually trying to answer.