LogInThe project is a reimplementation of DReCon, Ubisoft's 2019 character controller made using deep reinforcement learning.
Software Engineer and Programmer Ali Bharwani, a.k.a. Essential NPC, has recently shared a collection of demos, showcasing a fully physically-simulated humanoid model set up in Unity.
According to the creator, the project serves as a reimplementation of Ubisoft's 2019 DReCon, a character controller created using deep reinforcement learning, for user controlled physically-simulated characters using Unity-ML Agents.
"The difference between this and most other physically simulated controllers is that here the program actually controls every muscle in order to match the target pose," commented the developer. "It's not using IK to determine where a limb should be and moving it there, the neural net backing this has to learn how to 'flex' all of its muscles to achieve the target pose."
You can view the project's code here and download the demo scene over here. Also, don't forget to join our 80 Level Talent platform and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.