U.S. Patent No. 10,765,944: Machine learning models for implementing animation actions

Issued September 8, 2020, to Electronic Arts Inc.
Filed/Priority to September 7, 2018

If only the in-game bot was smarter… uh, oh…


U.S. Patent No. 10,765,944 (the ‘944 patent) relates to real time (or near real time) machine learning animation prediction. The ‘944 patent details an animation development system which analyzes game events, while predicting the outcomes of the events, to train better prediction of future events. It identifies an event during play of a video game, determines parameter values associated and a first outcome for the event, generates training data, and then generates and deploys a predictive model to make a prediction of a second outcome for a second event. One possible method of parameter values is to identify two objects and determine when the objects are in contact. An example of an event that could occur could be an animated character falling down. The ‘944 patent can be seen implemented in many of Electronic Art’s sports games.



An animation system and method generates predictive models that are deployed in animations, such as an animation associated with a video game, to predict outcomes resulting from events occurring in the animation, such as interactions between two or more objects in the animation. These predictive models may be generated based at least in part on training data that is generated by running the animation, such as playing a video game, generating parameter values associated with events in the animation, and determining an outcome of the events. The training data may be used to generate predictive models, such as by using machine learning algorithms. The predictive models may then be deployed in the animation, such as in a video game, to make real time or near real-time predictions of outcomes based at least in part on events in the animation.


Illustrative Claim:

  1. An animation development system, comprising: one or more processors; and one or more computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to: identify a first event that transpires during a video game play of a video game; determine one or more parameter values associated with the first event; identify a first outcome associated with the first event; generate training data based at least in part on the one or more para meter values and the first outcome; and generate a predictive model based at least in part on the training data; and deploy the predictive model in the video game to make a prediction of a second outcome associated with a second event that transpires during game play of the video game.