Adeia filed 5 patents this quarter across 3 categories: Cloud Gaming (2), AI & Machine Learning (1), and Streaming (2).
The Cloud Gaming patents focus on ultra-low-latency video encoding through parallel encoder instances and deep learning, plus technology enabling users to scan any screen and join a game session at the exact state shown in a livestream. The AI & Machine Learning filing covers a framework that uses time-aware deep learning to fit gaming sessions into user-defined time windows while maintaining consistent difficulty. The Streaming patents address real-time tutorial recommendations based on player performance metrics and machine learning approaches to optimize bitrate and quality tradeoffs.
Two cloud gaming patents tackle latency through parallel encoding architectures that leverage deep learning to predict optimal encoding parameters before frames arrive. Rather than relying on traditional trial-and-error rate control, both systems use variational autoencoders to analyze scene complexity and determine how many parallel encoder instances to deploy, each operating with different quality settings. The encoders run simultaneously, frame-synced, and the system selects the best output that fits within bitrate constraints, eliminating the need for re-encoding when conditions change.
A single AI & Machine Learning patent addresses a persistent problem in games with AI opponents: how to fit a session into a fixed time window without abruptly switching difficulty or swapping AI models mid-game. The system integrates time constraints directly into the Monte Carlo tree search algorithm, using dual neural networks that optimize simultaneously for game outcome and session duration. This maintains consistent opponent difficulty throughout the entire play window rather than ramping down artificially as time runs out.
Streaming applications receive 2 patents focused on contextual gameplay assistance and frictionless session joining. One system monitors player performance in real time and automatically surfaces tutorial videos matched to specific struggles, stitching together relevant segments and pruning outdated content when players improve beyond previous benchmarks. The other enables viewers to scan any screen showing a livestream and instantly jump into that exact game state without logging in or syncing accounts, using embedded metadata to reconstruct the session on demand.
All data sourced from USPTO patent filings. Google Patents may take several weeks to index recent publications. If a link is unavailable, search for the patent number at USPTO Patent Public Search.