Hello, VFC is something that's been on my radar for a while now. I would love to see this program used to assist in training intelligent action for real
I wanted to request an option to have a training dummy that at it's core was a playback device for mocap data. Ideally one that could become an active ragdoll when hit.
For example.
You enter the training zone. And there is an Unreal default rig you can "spar" with. The default rig will playback real mocap data of actual combos that the player would have to then Dodge and counter.
Hi Adam,
Thanks for reaching our!
Making VFC a practical tool for combat training is our goal.
The feature you requested actually sounds roughly like something we are already planning to do.
The plan of VFC's 'sparring' mode will happen sometimes into the future after our current workloads (The VFC Academy training basic toolset, and the single player career campagin mode),
The mode bascially like you suggested, detects and memorise player's performed action events such as punch type and head movement, and the timing of their occurance. And after a short delay playback the exact same movement (type) accoridingly.
For example:
Player enter sparring mode, start hitting the robot with a jab cross hook and so on. After 3 seconds, the robot will start to perform action that replicates player's action starting from the begininig (the jab cross hook).
This mode seeks to let player train in a spontaneous manner, where player can anticipate whats coming based on your own prior movement sequences, and react accordingly.
Hitting the robot will result in a ragdoll-like feedback, and if player can hit it hard enough while it was perfoming certain movement, it will 'disrupt' robot's current movement, which signaling a succesfully counter strike has been done, and congratulates player
Is this kind of what you're looking for?
Best
Xin
That plan is very ambitious, and actually closer to what I was thinking for an ideal vr fight trainer. They are all golden ideas.
The reason I mention mocap specifically is that the movements that VFC characters make are similar to real world sparring but still really glitchy, given that shoulders hips and feet have to be solved for by head and hand trackers alone. Which makes for an experience that isn't quite as as realistic as desired. If the bot were to have an animation system that procedurally made for realistically plausible fight animation like the UFC video games then the above concern would be a non-issue.
In my request I was essentially asking for a glorified animation playback system with a library of realistically animated combos to defend against because I had figured that practicing against a recording may be a relatively quick-and-easy-to-implement way of getting the "sparring" experience that trains the player to look for real-world telegraphs.
something like minute 2:00 of this video of the UFC 3 practice mode https://youtu.be/1s96Qpm3qzI
however with the ability to control playback speed. I believe a lot of effective training could be had very early on in development by using realistic recordings.
The Robot infinity mode has elements of this but being able to drill combos in the training room against bots at a constant rate would be invaluable. especially if you add the variation ideas you were talking about
But suffice to say that we are very confident that when it comes to animating the training bots, the animation will be tiers and tiers above the current PvP animations existing in VFC right now as it will be totally based on different set of system and foundations.
In PvP, as we can only detect player's head and 2 hands, rest of the body are really just a 'guess work' that we coded in, and hope that it translate as intuititvely as possible, this will subject to further improvement in the future of course, but as of right now, we are very aware of its imperfections.
As for training bot and also our next single player expansion contents, because now we are working with AI controlled humanoids instead of player controlled VR motion capture spots, we will now be able to use traditional animaton blending methods to present the actual animation, so it gives us the complete freedom to adjust and play around with the details of each animation, and parameter in which it varies.
A less obvious function that is very technical and hard to code is to let AI bots knows where to throw its strikes, so that factors in player's own recorded punch and its area of impact on Ai's body, and in turn AI will not only replicate the punch type and combo of player's action, but also the position in which those strikes were landed. e.g Player recorded punch is a head strike. then AI bot will be aim for the exact spot
And that calls for a very complexed animation sequencing and blending. But this is something we want to challenge ourselves to achieve. And we would also give player the option to adjust training bot's height and reach so the the player can experience all kinds of sparring scenarios.