Goals for this topic : Describing the inner workings of the module : highlighting the tricks used by Akiyuki for automating the robots' movements (without compromising his copyrights on the model and building instructions --sold at a very reasonable price, around 8EUR) Showing my mods and improvements on this module, and gather other people's ones and comments too, of course Like everybody, I was impressed when discovering this module for the first time. And although I partially guessed its working, I only discovered some mechanical tricks much later when watching the video again... and even a last one a few weeks ago when I built it ! That's why I want to share all of this. Official Akiyuki's video, as a reference : Now, (1) how it works (hidden for people who don't want to get spoiled, but prefer guessing)   (2) My mods and improvements Here is a bottom view of my build, where we can see the most : (a very very simple mod) Bringing the mechanical power from outside :
replacing the M motor with an axle which comes out at the right (sorry, hidden by the white beam supporting it) (mod) Spinning the robots in opposite directions :
replacing part of the central transmission unit (see Akiyuki's build manual p79-83) : central drive of both robot units made through a right angle gearing, see the 7L axle with Z12T gears at each end and a Z20T gear inside the fork. I also replaced the two Z16 input gears (below the 'L' beams) by Z20 ones.
I forgot to say that the robots must be mirror-built too (otherwise one robot will have its crankshaft spinning in the wrong direction, and its front being pushed downwards instead of upwards when it grabs a ball).
(for information, the original transmission has the Z20T gear parallel to the robot units' bases, driving them through a parallel gear train ==> odd number of gears from one robot unit to the other ==> same direction) (mod & improvement) Robot's position auto-reset if it loses its ball :
When the robot begins to spin with a ball in its hands (stage 3), it holds it too weakly because it is too light for the roller to roll down the tilted platform's slope and maintain the hands firmly enough. Although Akiyuki saw and tried to fix this problem by adding adjustable friction to the spinning of the robots' bodies, this makes the problem even worse when a robot drops a ball while in stage 3 (or sometimes when the robot exits its fast spinning phase = stage 4, without its roller falling into the well) because it stays in its position forever without human intervention.
That's why I replaced the adjustable friction generators with a detent mechanism which forces the robot to return to its home position (or to hold the ball more firmly), see the Z12 pinions at the rear, the partially hidden Z24 gears (with rollers mounted on opposite round holes) and the blue angled 4/4 liftarms acting as pawls.

My initial try was a simple crankshaft with a spring-loaded piston, but the shaft's torque around the home position was too weak, whereas it was too high around 90°(applying too much force on the robot's hands mechanism which has a high flexibility and backlash, so that it skips a turn, making the robot throw the ball away and returning home).
The alternate solution is a cam with the correct shape... which doesn't exist or can't be built in an enough small size to fit in the remaining room. And the solution came by dividing the robot's body rotation by two, hence a bi-foil cam looking like a technic beam (here two rollers in order to decrease wear) ! For more clarity, I can make 3D mockups if needed.