Skip to main content

Curious about neural networks and character control in video games? We've got a seminar for that!

Join us for a seminar on Phase-Functioned Neural Networks for Character Control, presented by Dr. Daniel Holden, UBISOFT             

Monday, May 29, 2017, 10:30am                                                  
EV  3.309


In this talk I present a character control mechanism for games which uses a new neural network structure called a Phase-Functioned Neural Network. In this network the weights are not stored directly, but computed via an additional function which uses the phase of the motion as an input.

Along with the phase, our system takes as input user controls, the previous state of the character, the geometry of the scene, and automatically produces high quality motions such as walking and running over rough terrain, climbing over large rocks, jumping over obstacles, and crouching under low ceilings.

Once trained, our system is also extremely fast and compact, requiring only milliseconds of execution time and a few megabytes of memory, even when trained on gigabytes of motion data.



Back to top

© Concordia University