Delayed-Light Rendering for Superluminal Objects

A graphics paper introduces delayed-light rendering for scenes where objects move faster than the signal speed used for imaging, while the simulation itself remains non-relativistic. Instead of faking the effect, the method enumerates visible images as roots of an emission condition against stored state history, so the renderer can recover multiple simultaneous appearances of the same body.

The interesting part for engine developers is that the solve is built around fixed-step history, where each segment becomes a quadratic. The discriminant identifies when image pairs are created or destroyed, and the slope determines playback direction, rate, and brightness. That makes the effect mathematically explicit rather than a special-case animation trick.

The system also builds a global perceived state from multiple observation events, using the upper envelope of their backward light cones. A monotonic emission clamp prevents the view from regressing to older images, and per-vertex solves handle bodies crossing delay gradients. Pre-generated frame-sequence assets can be treated as $(x,y,t)$ volumes sliced by the solved emission surface, which means reversal and de-phasing fall out without animation-specific code.

The implementation is planar in scope, so the visibility and occlusion problems of a full 3D renderer are left aside. Even so, the method has already been deployed in a released real-time strategy game, and the paper includes the optimizations needed to keep it interactive along with independent reference measurements.