Amazon.com, Inc.
Predictive modeling for multiplayer gaming

Abstract:

Machine learning is used to generate and update predictive models configured to predict the positions, orientations, or actions of various gameplay elements capable of being controlled by players of a network-based multiplayer game. A gaming server can send a trained predictive model to each player device actively involved in a gaming session. Those devices can use the predictive model to predict the positions and orientations of various gameplay elements, where the actual data will not be received for a period of time due to latency of transmission over at least one network connection. When the actual data is received, the game state and gameplay action can be corrected or initiated as appropriate, enabling the overall gameplay experience to be smoother and more enjoyable. The gameplay and prediction data can be analyzed to further train or improve the models.