Procedural Running / Galloping in Minecraft — Transcript

Full transcript

1. 0:03Hi. This is a followup to my last video  about procedural walking animations.
2. 0:09Since then, I added support  for different body plans.
3. 0:12Here's a really small spider.
4. 0:15And here's a really big one.
5. 0:18Here's a quadruped.
6. 0:20A hexapod.
7. 0:22And an octopod.
8. 0:26They all work using the same ruleset  I described in the last video.
9. 0:30All I had to do was generalize the functions I  used to group adjacent and diagonal leg pairs.
10. 0:37I made a few improvements to the ruleset.
11. 0:40The legs now move ahead of the rest position.
12. 0:45I also added an adjustable cooldown to space  out the movement of diagonal leg pairs.
13. 1:01Next, I added galloping.
14. 1:03While galloping, we group legs into  horizontal pairs. A leg in each pair
15. 1:08is designated as the dominant leg. When the  dominant leg exits the trigger-threshold,
16. 1:15we will move both legs at the same time.
17. 1:22We can add a cooldown between vertical  pairs to prevent that leap-frogging motion.
18. 1:37We can also customize the gait by adding  a cooldown between horizontal pairs.
19. 1:54Unfortunately, this does not play  well with our current gravity system.
20. 2:02In the last video, we apply a normal-force  only when there is at least one grounded
21. 2:07diagonal leg pair, but this is too big  of a simplification to support galloping.
22. 2:12So I came up with a more elaborate rule set that  doesn’t require a full rigid body simulation.
23. 2:17The first step is to specify a center of  mass. It does not need to be accurate,
24. 2:22as long as it's in a believable position.
25. 2:25Next, we will draw a polygon  around each grounded leg.
26. 2:29If the center of mass is  contained within the polygon,
27. 2:32we apply an equal but opposite force to gravity.
28. 2:36Otherwise, we apply the force at an angle,  originating from the closest point on the polygon.
29. 2:50To detect if a point is contained in a  polygon, we will cast a line from the
30. 2:55point towards an arbitrary direction. If  there is an odd number of intersections,
31. 3:01then we know it is contained in the polygon.
32. 3:09An unintended benefit of this new system is that
33. 3:13the spider now falls off ledges  in a much more convincing manner.
34. 3:49I also added cloaking.  Each segment locates a block in the
35. 3:54player's line of sight and camouflages  itself by adopting a similar texture.
36. 4:00However, it does not copy the block directly.
37. 4:03For one, it avoids switching to blocks  with directional textures. As an example,
38. 4:08grass blocks are mapped to moss blocks.
39. 4:11For another, it also avoids transparent  or non-full blocks. Leaves are mapped to
40. 4:15moss blocks, while slabs and stairs are  mapped to their full-block counterparts.
41. 4:21I thought the camouflage pattern worked too  well. To add a bit of imperfection , I made
42. 4:26it intermittently select a different block with  similar color. The randomizer is seeded with
43. 4:32the segment's position, that way it still  looks random without being too chaotic.
44. 4:37For the glitch effect, I use glazed terracotta.
45. 4:42To address some of the questions I got in my  last video, here's a more detailed visualization
46. 4:48of my inverse kinematics implementation.
47. 4:51A chain is composed of a root position, and any number of segments, each containing a length and a tip-position.
48. 5:00To move the chain to a new position, we first  set the last segment's position to the target.
49. 5:06Then, we will subtract the parent's position from
50. 5:08the child. This will give us a  vector indicated by this arrow.
51. 5:13Next, we will normalize the vector.  We will divide it by its own length,
52. 5:17resulting in a new vector with a length of 1.
53. 5:20Finally, we will multiply the  vector by the segment length.
54. 5:24This vector now represents  the segment's base position,
55. 5:27which we can use as the  target for the next segment.
56. 5:31We will move the chain back and forth,
57. 5:32until we find a solution within  some acceptable margin of error.