Videos

Phase-Functioned Neural Networks for Character Control



Yoshiboy2

We present a real-time character control mechanism using a novel neural network architecture called a Phase-Functioned Neural Network. In this network structure, the weights are computed via a cyclic function which uses the phase 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 that achieve the desired user control. The entire network is trained in an end-to-end fashion on a large dataset composed of locomotion such as walking, running, jumping, and climbing movements fitted into virtual environments. Our system can therefore automatically produce motions where the character adapts to different geometric environments such as walking and running over rough terrain, climbing over large rocks, jumping over obstacles, and crouching under low ceilings. Our network architecture produces higher quality results than time-series autoregressive models such as LSTMs as it deals explicitly with the latent variable of motion relating to the phase. 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. Our work is most appropriate for controlling characters in interactive scenes such as computer games and virtual reality systems.

Source

Similar Posts

48 thoughts on “Phase-Functioned Neural Networks for Character Control
  1. For your functions is n the number of points/arrows on the line? Be cool to apply that to rpgs as I would assume the character would move better over complex terrain the higher the value of n so it would be fun to have it in an rpg. If their say agility, dexterity, sense and or strength were lower than normal that would lead to a lower n meaning they wouldn't be able traverse complex terrain well or fight on it. The complexity you could add to games like planescape, baulder's gate, pillar of eternity and more would be amazing. It would be hilarious to see a mage cast a spell like fireball and when they go to throw it at the enemy they stumble on the terrain and hit a teammate or blow themselves up.

  2. Very good. I see some of the comments are about Uncharted. Alas Uncharted is but a cheap imitation to that of neural networking the movement of a character. Now all they have to do is give him his own ai, make him a self aware NPC, and and see how long it takes him to realise he is in a computer simulation. I'll see myself out. ?

  3. this is still just a dynamic animation system, right? the movement is not being produced by real-time physics calculations and simulated muscle and joint movement? still impressive but i hope eventually games are able to do the latter

  4. Do you guys happen to have a research paper on any link to something? I'm very interested in playing around with this tech and I don't know where to start.

  5. does anyone what happened to them? are there already games using this technique or are planning to use it?

  6. Ooooo I like that idea, where you look is where you're character looks and their walking is different, like walking backwards and sideways while facing your target of vision

  7. I hope the guys from @InfinityWard and @Activision have noticed this and contacted you. This is superb stuff!

    (oh, and I just shared this vid with them on twitter)

  8. Truly outstanding work. Congratulations on the fantastic results and thank you for all of the hard work. I'm always thrilled to see new steps forward (no pun intended) in software development that cuts down on some of the tedious tasks of content creation.

  9. This is so splendid. Such an impressive neural network. I would be interested to read/learn more about the phasing mechanism – it gives such mind blowing and smooth results!

Comments are closed.

WP2Social Auto Publish Powered By : XYZScripts.com