[go: up one dir, main page]

A downloadable game for Windows and macOS

A 1v1 competitive volleyball game with an AI opponent made using Unity ML-Agents and reinforcement learning.

Controls

  • Z / A BUTTON on Xbox controller - Jump, Dash while mid-air
  • X / X BUTTON on Xbox controller - Hit the ball
  • ARROW KEYS / LEFT STICK on Xbox controller - Move, Aim the dash

Tips

  • You can press the hit button before the ball is within radius and still hit the ball - the hitbox stays active for a few frames after the initial button input.
  • You can only hit the ball 3 times consecutively. Using all three touches before hitting the ball over is a good way to set yourself up for spikes.
  • Dashing up to block the ball from coming over to your side of the court is a great aggressive strategy.

Made as a Bachelor's Thesis Project at University of Tartu.

Download

Download
Voll-AI.zip 24 MB
Download
VOLL-AI - MAC.zip 24 MB

Comments

Log in with itch.io to leave a comment.

(+1)

Machine Learning is cool

(+1)

pls change control

Hello! Would you mind telling me what exactly you didn't enjoy about the controls? How do you feel they should be changed?

(+2)

I agree with the controls being a bit difficult, esp on keyboard. Maybe arrow key up to jump, z to dash and x to throw would be good. With the current setup I have to think of 3 different inputs just to do a simple jump and throw. Adding some grace to the jump (when you input before landing that it counts to the next jump) would be good as well. Perhaps having a stronger visual indication when the ball enters your reach would also be nice.

Awesome game, though! The AI felt very human-like to play against. Good luck with the thesis!

Thank you so much for the feedback! I really appreciate it. 
It is a bit worrying that the controls are not very intuitive, because this probably affects the AI's win-rate by a lot in their favor. The AI may seem more intelligence than it actually is because of this.

If I may ask further - do you feel that the issue is solely in the controls or is the inherit mechanic of how the player interacts with the ball hard to control/play with? I hope to revisit the controls some time after the thesis is complete, so the game could be more enjoyable as a "game" and less as just an "experimental thesis project".

(+1)

Hey! It's hard to say, but it's mostly just a lack of feedback and not a lot of room for error I think. Maybe helping the player along in some parts. (eg: slow the ball down when it's in range, adjust the aim a bit in your favour, etc) to make it less precise and more fun :)