Activision Blizzard

The sole Graphics patent addresses a problem that has long demanded significant manual effort from artists: generating level of detail meshes for real-time rendering. The system automates this process by deploying virtual imaging probes arranged in spherical and hemispherical configurations throughout a scene, using GPU-accelerated processing to determine where those probes should be placed for maximum effectiveness. The result is a pipeline that produces meshes with fewer triangles than conventional grid clip-map methods, reducing the burden on artists while still optimizing rendering performance across a range of Hardware capabilities.

4 Game Engines patents cover a broad spread of production and gameplay challenges, touching everything from content delivery to animation quality. One system allows a game to dynamically assemble a shorter, time-bounded session from existing content in real-time, automatically adjusting narrative flow and equipment loadouts so the experience feels coherent rather than truncated. A second patent describes a motion capture setup built around large LED walls that track each actor's individual field of view simultaneously, rendering a personalized virtual game environment for every performer at once, something film-oriented LED volume systems do not do. The third patent covers an automated testing framework that scripts avatar movement along consistent paths through open-world environments and aggregates performance data from multiple Hardware configurations running in parallel. Rounding out the category, a fourth patent tackles first-person weapon animation during corner slicing in FPS games, using an edge-scoring algorithm that weighs geometric factors like the player's forward vector and edge normals, then applies LERP interpolation to produce fluid weapon movement with separate parameters for hip-fire and ADS modes.

The 1 AI & Machine Learning patent describes a system in which NPCs are trained on granular behavioral telemetry captured from real players, allowing the resulting bots to replicate specific individuals' tendencies, skill levels, and playstyles rather than defaulting to generic scripted behavior. Because the underlying ML models are designed to work across Platformss and multiple game genres, a single generation engine can produce bots that mirror a friend, a professional player, or any other human whose data the system has ingested, and those bots continue updating as the players they replicate evolve over time.