EA

The 3 AI and machine learning patents reflect a shift toward more naturalistic and generative systems. EA's sensory awareness system gives non-player characters a perception-bounded decision model, meaning they respond only to stimuli their virtual senses could plausibly detect rather than drawing on omniscient game state data. A separate filing describes a variational autoencoder that produces character animations on the fly by sampling from learned pose distributions, removing the need to load or query motion capture libraries during play. The third patent tackles procedural level design through a neural network that accepts designer-specified constraints like difficulty, symmetry, and size, then generates layouts matching those parameters by conditioning both encoder and decoder on structured feature inputs.

EA's game engine patent proposes an architecture where simulation and rendering operate independently by passing state data through a Streaming interface. The presentation layer interpolates between snapshots to generate frames without requiring the simulation to advance, decoupling visual refresh rate from gameplay logic tick rate. Multiple simulation threads can also run at different speeds in parallel, each feeding the renderer on its own schedule.