Show HN: Quake 1 movement physics reinforcement learning project

Re:Show HN: Quake 1 movement physics reinforcement learning ... by fdalvi on Sunday May 03, 2020 @04:21AM (#23057684)

It wouldn't be difficult at all if the optimal running technique was known before hand; I think the goal of many of these RL exercises is to either i) find a better solution than what we may have imagined or ii) confirm that our knowledge was indeed the best possible solution!

[ Parent | on HN ]

Re:Show HN: Quake 1 movement physics reinforcement learning ... by bnegreve on Sunday May 03, 2020 @05:31AM (#23057929)

> It wouldn't be difficult at all if the optimal running technique was known before hand;

You're right, but this is a simulated environment. It means the physics in it are driven by a small set of known (and most likely deterministic) rules.

So I wonder if it would it be possible to analytically determine what is the best possible running strategy.

[ Parent | on HN ]

Re:Show HN: Quake 1 movement physics reinforcement learning ... by kipi on Sunday May 03, 2020 @04:31AM (#23057716)

Correct, it wouldn't be hard to take the best known strategy and implement it perfectly. Aside from the two points you mention, it also serves as a precursor to applying RL to real levels in which we may not be so sure about the optimal strategy.

[ Parent | on HN ]