PhysX, inverse kinematics for robotics & Co.

Before specific development on R.C. Bot Inc. started, Blue Budgie Studios tested Unreal Engine and PhysX extensively to make sure it can be used for gameplay. Two test projects were started: a minor project on passive ragdolling, basically throwing ragdolls around, and a major project on active ragdolling, i.e. biped walking and self balancing using only motors and rigid bodies without any animations, all controlled by an artifical intelligence. Results from this project are shown in the video on this page. The project was stopped when first steps of the ragdoll were achieved. Well, steps might be the wrong word. Asimo by Honda does a better job. But still... biped walking, the actual walking, not just jumping and stuff, is really difficult 100% physics based. The jump at 2:20 in the video below is nice though. Euphoria by Natural Motion actually pushes such things to the extreme.

Or this here:

https://www.youtube.com/watch?v=pgaEE27nsQw

http://www.goatstream.com/research/thesis/index.html

 

Active Ragdolling Tests (robot fully assembled with PhysX joint, all actions driven by PhysX joint motors)

 

Passive Ragdolling Tests (Arms/Legs moved by forces pulling on hands/feet to get ragdoll up)

Here is some background on Inverse Kinematics:

Most robotics applications move limb-like things around, may it be for industrial robots or future technologies like biped walking. Take for example a robot hand that reaches to a door knob. Using this example, one would have to move the shoulder and the elbow to certain angles in order to move the hand to the door knob. How do you know the angles? You could try first. Say shoulder 45°, then elbow 10°. That did not work. The hand is not at the door knob. What next? Try again? No! The solution is inverse kinematics. It tells you the shoulder angle and elbow angle to reach the door knob. It does that by solving a set of equations, where you input the door knob position and the shoulder position. Door knob, elbow and shoulder are said to be the inverse kinematics chain, and in this case one would speak of 2-bone inverse kinematics chain - your arm has two major bones.

Two bone inverse kinematics can be solved analytically with relative ease, whereas higher order chains with more bones require matrix solutions, which most often are solved numerically. It should be noted, that two bone chains can be stacked to higher order chains if one is willing to do without certain degrees of freedom in the final chain.

R.C. Bot Inc. uses inverse kinematics heavily to control the legs of the robot in every situation. It makes sure, in conjuction with active foot placement, that gameplay can be achieved. Especially foot placement is an important aspect to make sure the robot adapts to the underlying ground all the time. All inverse kinematics and foot placement in R.C. Bot Inc. is custom made from scratch by Blue Budgie Studios.

And here some impressive real world example

Druckversion | Sitemap
© Blue Budgie Studios