Controller Renegade Project [DriverN Shooter system]

aiWalker evade variations (cont'd)
red is enemy, green is destination


+ look at enemy

 

here's how loop can be done:
first target set, and then usin seq of target-change based on distance
three loops but stops only on the next


there's ASeq () for anim seq too

 

adding Ezy-Slice

 

the different versions going from ragCopy:
simple going ragdoll

going animated

and going blended

 

I'll add this little script I use when want to redo the T-Pose
(must be removed from project, will not build)
it exports to .asset, which then can be exported as .obj

Code (CSharp):

1. using System.Collections;
2. using System.Collections.Generic;
3. using UnityEngine;
4. using UnityEditor;
5.  
6. public class exportMesh : MonoBehaviour {
7.  
8.     public SkinnedMeshRenderer[] smr;
9.     public string folder;
10.     public KeyCode export = KeyCode.Space;
11.  
12.     void Update () {
13.  
14.         if (Input.GetKeyDown (export)) ExportMeshes ();
15.     }
16.  
17.     void ExportMeshes () {
18.  
19.         AssetDatabase.CreateFolder("Assets/export", folder);
20.  
21.         for(int i=0; i<smr.Length; i++){
22.                            
23.  
24.             Mesh msh = new Mesh();
25.             smr[i].BakeMesh(msh);
26.            
27.             AssetDatabase.CreateAsset(msh, "Assets/export/" + folder + "/" +smr[i].name + ".asset");
28.         }
29.     }
30. }

 

the automatic jump works now

next will do some method to choose jump style based on some condition
(not just the first one that comes up)

noticed that for not-forward jumps the raycast needs to be rotated
[raycast at jumpPoint is used to check if hand-support is till there]

so added variable for angle
public float jPCastAngle;//raycast angle at jump point [default 0 is forward]
otherwise would not detect that jump is okey
and would go ragdoll

 

ragCopy is called ragBlend now
and integrated in Renegade

 

locostart
[animations not included unfortunately]



all controlled by 4 arrows

 

upbodyMX probably will be the primary solution for gun animations
though weapon state machines can be used too

can create and modify weapon animation
separate weapon locomotion

benefits
- no need for even the one leftHand IK (weapon grip), if the pose is precise
- breathing unaffected
- no need for animations (lower file size)

drawbacks
- cannot do load/draw weapon anim (though might be possible to use one)

 

also a punching bag demo that wasn't showcased

 

improved strayLoco


also started to redo the sword mechanics
to be based upon (start from) left and right stance
something like this [preview]

so if the move changes stance, no need to define
if it is a left cut or a right cut (because the stance is tracked)
but there are slashes which do not change stance
and haven't decided what to do with those

 

with this simple script was able to correct one cmu animation
used along with playTime

Code (CSharp):

1. public class fixerBlend4BoneX : MonoBehaviour {
2.     //for animator >> based on normalized time
3.     //blending between two vec3s w animCurve [of normTime]
4.     public Transform bone;
5.     public Animator animator;
6.     int i, x;
7.  
8.     public float nTime;
9.     AnimatorStateInfo animStateInfo;
10.  
11.     [System.Serializable]
12.     public class rX {
13.        
14.         public Vector3 rStart;
15.         public Vector3 rEnd;
16.         public float tStart = 0.5f;//[normTime]
17.         public float tEnd = 0.5f;
18.         public AnimationCurve animCurve;
19.     }
20.     public rX[] RX;
21.  
22.     public float perc;
23.     public bool overridePerc;
24.     [Range(0.0f, 1.0f)] public float percOverride;//for manual realtime check
25.  
26.     public bool multi;//rotation multiplied or straight
27.     public Vector3 rFinal;
28.  
29.     [Space(10)]
30.     public bool rotation = true;
31.     public bool position;
32.     bool inRange;
33.  
34.     void LateUpdate () {
35.  
36.         animStateInfo = animator.GetCurrentAnimatorStateInfo (0);
37.         nTime = animStateInfo.normalizedTime;
38.  
39.         inRange = false;
40.         for (i = 0; i < RX.Length; i++) {
41.  
42.             if (nTime >= RX [i].tStart && nTime <= RX [i].tEnd) {
43.                 x = i;
44.                 inRange = true;
45.             }
46.             if(nTime > 1 && RX [i].tEnd == 1) inRange = true;
47.         }
48.  
49.         perc = RX[x].animCurve.Evaluate (nTime);
50.         if(!overridePerc) rFinal = Vector3.Lerp (RX[x].rStart, RX[x].rEnd, perc);
51.         if (overridePerc) rFinal = Vector3.Lerp (RX[x].rStart, RX[x].rEnd, percOverride);
52.  
53.         if (inRange) {
54.             if (position) bone.position = rFinal;
55.             if (rotation) {
56.                 if (multi) bone.localRotation *= Quaternion.Euler (rFinal);
57.                 if (!multi) bone.localRotation = Quaternion.Euler (rFinal);
58.             }
59.         } else {
60.             if(!overridePerc) rFinal = new Vector3(0,0,0);
61.         }
62.     }
63. }
64.  

the original on the left

and with one more clean-up

 

improved helicopter ai

a helicopter shootout mini-game
https://www.mediafire.com/file/kxgm2wd2fytxqfm/heliShootout.ZIP/file
was late to realize that there's a high propability the enemy copters crashing into each other
because they are crossing their paths [will switch them next time]

controls:
fire-up leftAlt
accel leftShit
slow leftCtrl
← ↑ ↓ → J G

 

adding uGuiTweenTool

 

about the sword mechanics

so issue was that a cut from left can only be started from left stance
so had to do switch stance before cut

had all kinds of problems with triggers, events
even started to redo it into based on state normalized time
but went back to it, now kinda works


here you can see couple frames where the attack should have started sooner
it's not the code (the anim should fire) there is dark zone
but crossfade should work even while transition in mecanim

 

before it gets forgotten and redone
snapshots of upbodyMX and pointAt

pointAt is kinda final : spine and head lookat had been added

the execution order is (if I'm not mistaken)
- ikControl
- Renegade.ragBlend
- Renegade.CCD
- upbodyMX
- pointAt
- fingers

contrary to this, ragblend cannot be initiated while pointAt is on (weird results)
after fixed and copying ragdoll, pointAt can be turned on

 

simple smoke turbulance

 

Renegade.CCD.head
locoStart false

 

quick note on strafe+upbody:
hip should look right, like the base pose (left foot forward)
here going left is left foot forward, going right is right foot forward


---
now fingers can use quaternion and eulerangles too
play/copy back and forth

---
animator test results
performance of 100 actor models
animator: one layer
actor:
1 obj 6500 tris: 8.0ms 125fps
2 obj 1500 tris: 8.5ms 115fps
2 obj 6500 tris: 9.5ms 105fps

so it's very much worth to combine and use models made of one object

 

looks like this could work

 

this is sad

5.6

2019.4

 

the right method might have been found

this stance-based approach is going to be implemented to fighting too
the previous #265 I think was my mistake initiating crossfade to same state

-there's events for every attack: sAttackOn sAttackOff stanceL/stanceR
-needs: root rotation baked


-can be two stances for one attack
startin from R: going to L and ends on R

 

as you may know, assetstore doesn't allow uploadin projects with pre-2017 version now.
so the 5.6 scenes are going to be included zipped

the latest upload was rejected (probably for issues w 2018.4, though the code gives no errors)
so gonna use this time to go for the changes wanted to do for optimization sake
- not using gravityWeight anim curves
- limiting the use of statebehaviors (which doesn't update well for 2018.4)
- using hashes instead of strings for crossfade [done]
- using CrossFadeInFixedTime [done]

and a thing noticed just now
CrossFadeInFixedTime can transition to the same state, while CrossFade cannot!

this has a big effect on animator and code
although since there is animSelect on blendtree, must keep this version
unless redoin param float lerp


it's better to switch off dynamic capsule height for crouch
and use curve on animation for it

update
aiWalkerMX: stuck detection (by velocity and by position) > bool isStuck
jumpMX: switchin off capsule while targetmatching > bool switchOffCapsMatchT
capsule height: for every state (loco, slide, crouch, jump etc) could be now
1. dynamic capsule height
2. capsule height by curve on animation
3. hard-set value for capsule height

 

aiWalkerMX:
avoidance + auto jump

 

new feature idea:

virtual transition list

next action list (blend tree) reusable for MX
initiated by nTime/event+input if ready
every action prop can be array
(and can be by selected float, with orderingMind: analyzed next by btmatrix)
so its basically crossfadein in a mecanim state (seq) onjumpx, onpunchx, onkickx, onslidex, onrollx...

state : second move
punch : avoid elbow knee (backfliponknee roll sweep... )
kick : dodge kapuera elbow spin backstans jump flipjump...
slide : roll getup ...
jump : ground headground crouch spin...

default third: getup

CrossFadeInFixed

class actions
bool[] ready
bool[] ins
hash[] action

class order
float[] nTime
int[] pose
btmatrix[pose][action]

should look like this:
(animation seq test)

 

body pose analyzer


roll by pose

 

okay, finally got this workin (for roll and slide)
there's secAction based on bool use and int pose, initiated on contactPoint

root transform rotation should be baked
because if it's not, can mess up pose detection
see 1. baked 2. non-baked



can be autom or can be input
input should be pressed before contactPoint

looks like this
without secAction

with secAction.slide + another anim, second roll

with secAction.roll + another roll


and there's weight for slide/roll (higher one is played)

 

doublejump working alright

also added switchAction
here's how it looks w 0.2 and 1.0 blend value
more blendin helped to look better (but need to calc in com n vel)




wanted it able to play any state on mecanim
so it needs a bit more configuration for any state
it checks for pose and receives input in jumpstate ungrounded and that's it, plays exe


probably will throw out Renegade.state and pose check for jumpin initiation
ground check is enough, haven't had to use them yet, maybe velocity check should be+
probably reversed my initial plan, cause I think that if even possible probably d'be more complicated
would that be anystate config matrix and this is more like estimated chaos : ),

so for now
there's slide roll jump jumps jumpOn jumpOver jumpDwn on JumpMX2
and jumpClimb should come next

jumpx : the initiated jump
doublejump : new jump while jumpstate grounded
onJumpx : reachin while jump > ragBlend
secAction : changing jumpstate on contactPoint
switchAction : changing jumstate ungrounded

 

added groundcheck event
and stumble, fall-pose > ragdoll behavs
doublejump animstart

 

jumpClimb first version workin
with IK handpositioning

the blendttree transition looks like this

ingame

 

added second matchtarget for jumpClimb
added targetP offset
added rootP adjustment by animationcurve (see actor on the left)

jumpClimb > doublejump

 

think I forgot about this one
got buildgen kinda work with floors but should redo it

builds generated


needed jump > freehang animations
[because transition issues w rootpos/com]

 

customers are eligible for RawFormula boards access.
drop me a line on pm w a proof of purchase.
new content coming [game ready models]

try to get up a new version [0.98] on the store as soon as possible

newest versions of scripts, prototype packages available upon request

 

couple issues will have to chase down on this too


http://www.mediafire.com/file/ccgcqqg5rqdgh7x/aiFighter001b.ZIP/file

set fade times to 0.15
and animator state speeds to 1.15

the keyboard controls a bit awkward because of key-combos
gamepad should feel better

fight has different loco and steppin states
can mix but if key-input not on time looks bad [fwdArrow / Tkey]

the stumble feature on JumpMX
but sometime its too much

and when doesn't stumble

 

added sounds
added anim switcher
ins.LocoReset : key L
added IK attackStop
http://www.mediafire.com/file/1rwrcf5mrtm6xp5/aiFighter001D.ZIP/file

and there's some anims available on mi patreon page

 

blended result

 

added muscle curve modifier tool for extracted curves


and basic rootmotion P and R tools

 

added new variations for fighter movement


can rotate


after I build the new fighter demo will start to construct the scenes for the new version and send it in

 

attention customers:
got a complaint from someone, that project downloaded from the asset store had missing files,
so I sent him the missing files.
if you have similar issue, contact me.

 

fighter demo:
added GP control and some improvements
still has some issues, but way better what started with

http://www.mediafire.com/file/h0zyi40ojxew85q/aiFighter001E.ZIP/file

also tested out 80 Renegade actors in a scene
and runs pretty good on my comp
http://www.mediafire.com/file/zcqxif6n5ieabqd/test80.ZIP/file

 

already started to work on 0.99 version
I'm going to tear it apart and build it up again

going to remove 2D blendtrees.
::the usual way to do 2D blend trees is bad because
- if idle anim is at center (0,0), when input blends through that value it adds noise to the animation (only can make it less noticeable, if idle pose looks like movement pose)
- and also abrupt changes in blend parameter cause root position error
(that's why can't do direct walkF > walkB / walkL > walkR)

with the new design it is possible, no root position error and I think every feature of the 2D blend tree can be replicated
look at the prototype (the closer one)



http://www.mediafire.com/file/899azktz33gwc6x/R0991.ZIP/file
II pretty like it
it needed a couple tricks to work
and still needs a bit of coding yet


I'm going to switch to post as mi publisher account next, do nutt be surprised

 

the diagonals don';t have runs
the runs don';t have stops
but the fwd - back seems to work w speeds too

might be trackin leg fwd and startin new move based on that

the controls are the usual
G F : left / right
M N : speed up / down
http://www.mediafire.com/file/4rp23vm00zbkz1g/R0992.ZIP/file
update: walkB wasn't synchronised

 

http://www.mediafire.com/file/v75gvi6y43zvo0u/R0993.ZIP/file

added input blending on diagonal too
added knee position tracker
and stance detector (automatic doesn't detect neutral rootR)

http://www.mediafire.com/file/anmaoypvgidw7mh/R0995.ZIP/file
there's a bug with idle transition caused by animator parameter

 

added turns which can be interrupted (at any time by loco)

prob the idle transition issue still there just not visible, because input method was changed
there's the occasional root hickup I'm guessin the animator blend order is the culprit (must try separate all states' parameters)

A D : turn L R
W S : turnAngle
http://www.mediafire.com/file/xfykxqppexyr9uj/R0997.ZIP/file

all have to track those points


there's also hip, spine, hand positioners (ccd, anim.bodyPos) and feetIK in plannphase
I think for lowering head animations are better
for physics reactions and results can use hip-spine modif (+limbs limited)
and for avoidances (that with focus on fees)
for aiWalker not needed (can do all through animator)
might be good for transitional states like poses (falling pose, defending pose etc)

 

the update is in the asset store

uploaded with 2018.4 but includes 5.6 scenes

 

spine ccd at work
with right hand at animated position + offset


latest demo
http://www.mediafire.com/file/505sta09g50o5sv/R0998.ZIP/file

 

Hi @bobadi is this Template works on mobile and do they have mobile input control?
Thanks!

 

Hi

no, it doesn't have mobile input control
but it is easy to integrate because every input variable is separated in Renegade

like this
simplified example:

Code (CSharp):

1.     void Update () {
2.         if (useInput) checkInputs ();
3.     }
4.     void checkInputs() {
5.         h = Input.GetAxis ("Horizontal"); //float
6.         v = Input.GetAxis ("Vertical");
7.         ins.doRoll = Input.GetKeyDown (ins.roll); //bool
8.     }

you have a mobile input package, where you get float/bool values
in your mobile control script: float hx, bool rollx

don't have to anything with Renegade code,
you just make a controlscript (see RenegadeGP for reference)
you make useInput = false in Renegade
and you just replace input variables like this in your controlscript

Code (CSharp):

1. public Renegade Renegade;//drag and drop here
2. void Update () {
3.     Renegade.h = hx;
4.     Renegade.ins.doRoll = rollx;
5. }

and would probably want to add it before Renegade in script execution order

 

gonna include IK control inside Renegade from now
here's how Renegade looks now in inspector


also tested 32 actors and got 40 fps inside editor, so I'm happy with that.

http://www.mediafire.com/file/3008ryg3h98eki4/R0998_32.ZIP/file

there's a strange bug in it
first the turning animations work, then somehow as if there was no animation (it goes default binding pose)
oh I see, it's missing anims in the turning blendtree on high xc (speed input)

 

another animation tool

from this to that