[TC23] Stewart platform GBC

[Heppu]

Hi, first time attempting to attending an Eurobricks competition. We shall see how this goes since my plan *might* be a bit too ambitious. The idea is to have a EV3-based 3DOF stewart platform to move the balls. I've wanted to make a stewart platform out of Lego for a while but never had a good application for it. GBC seemed like a good fit.
 

Consept mechanism:

Spoiler

 

First consept to move the arms of the stewart platform was to use EV3 medium motors that turn turntables trough wormgears. This required quite a bit of vertical space and and turning speed was kinda slow. That is why I decided the attempt the same with large EV3 motors directly moving the arms, which is the current design Im working on. Wormgears do allow the mechanism to lock in place though, so if the platform ends up being heavy and sagging the arms under its weight, I might reconsider and visit this approach again.

 

Current status:

In a three degree of freedom Stewart platform, 3 large motors working in tandem can change the platforms angle in 2 axis (roll and pitch) as well as move it up and down (heave). The current plan is to have a free moving pendulum with "ball fork" that pivots around the center of the platform. This fork would pick up balls on one side of the machine and swing to drop them off on the other side. The system will of course need an input sorter to lift the balls to required height. It will most likely be a semi standard stepper module.

If I find the pendulum fork to be too unreliable or too slow to meet the 1ball/sec requirement, I might need to simplify the system. Most likely ditch the idea of the pendulum and use a fixed bucket and shute system attached to the the platform itself. There are still a lot of questionmarks with this project and wether it will finish in time. But I figured If I present it here, it might give me more motivation to work on it ;) ! Next up in line is to start on the code for controlling the platform and designing a better pendulum as well as some sort of input and output racks that mesh with it.

Rules violation

As is evident from the pictures my model includes non Lego elements. These are:

• Raspberry Pi with Brickpi3 board to control the EV3 motors.
• 3D printed case for the RaspBerry
• Custom made shorter length EV3 cables

I understand that these violate the rule: "All entries are to include only real LEGO. No clone brands, 3rd party parts, or digital entries allowed." Since these are not part of the core mechanical mechanism to move the balls, I'm hoping these would be allowed. If not, I'm totally ok with that and will happily continue this project outside the scope of the TC23 competition.

[Berthil]

Very interesting and will be fun to see the progress! In the GBC rules 3rd part PF motors and 3rd party cables are allowed as these are hard to get. I'm sure the cables and case is not a problem but may be @Jim can comment after his holiday.
As for the Raspberry Pi, why use that to control and not an EV3 or other Mindstorms unit?

[Heppu]

11 minutes ago, Berthil said:

As for the Raspberry Pi, why use that to control and not an EV3 or other Mindstorms unit?

Sadly I do not own any working EV3 or other mindstorms hubs.

I will use ev3dev operating system on it that supports both EV3 brick and Brickpi. There should be no major perceived advantage to using a raspberry. 

Edited August 30, 2022 by Heppu

[howitzer]

Very interesting, I'm looking forward to see this being built, for the contest or otherwise. I happen to have a spare RI hub, but sadly that's not compatible with your motors and probably too big of a hassle to lend for your project anyway... I hope you can find solution with Jim and others to make this build eligible for the contest!

[Jim]

On 8/30/2022 at 10:13 PM, Heppu said:

 

 

  Reveal hidden contents

 

First consept to move the arms of the stewart platform was to use EV3 medium motors that turn turntables trough wormgears. This required quite a bit of vertical space and and turning speed was kinda slow. That is why I decided the attempt the same with large EV3 motors directly moving the arms, which is the current design Im working on. Wormgears do allow the mechanism to lock in place though, so if the platform ends up being heavy and sagging the arms under its weight, I might reconsider and visit this approach again.

 

 

 

 

Rules violation

As is evident from the pictures my model includes non Lego elements. These are:

• Raspberry Pi with Brickpi3 board to control the EV3 motors.
• 3D printed case for the RaspBerry
• Custom made shorter length EV3 cables

I understand that these violate the rule: "All entries are to include only real LEGO. No clone brands, 3rd party parts, or digital entries allowed." Since these are not part of the core mechanical mechanism to move the balls, I'm hoping these would be allowed. If not, I'm totally ok with that and will happily continue this project outside the scope of the TC23 competition.

I’m okay with using a raspberry as a controller. The contest is about the contraption, so I can look the other way when it comes to controlling the motor. As long as there are no objections from Berthil or other members.

On 8/30/2022 at 10:32 PM, Heppu said:

I will use ev3dev operating system on it that supports both EV3 brick and Brickpi. There should be no major perceived advantage to using a raspberry. 

 

[Berthil]

No objections from my side 👍

[9v system]

no objections from me ether.

[Heppu]

Great to hear that everythig seems to be good to go ahead!

I finished most critical parts of the control system code. Most important I calculated the inverse kinematics for the platform. It works such that I can input the height of the platform, tilt angle and the direction of the tilt. These are then converted and sent to desired motor angle coordinates. Second I made a reset function that calibrates the motor angles to a known value. It drives all the arms 1by1 towards the base until they drive against the axle with red bushes and and become 'overloaded'.

Physically not much has change yet. The platform legs are changed from the modified bent liftarms to straight 7L liftarms to get a geometry that enables larger range of motion.  

 

made a small video to showcase the progress: In it the motors are calibrated and then the platform does 3 loops around at 10° angle.

I was hoping I could pick up balls by tilting the pendulum up from below the box. Seeing the range of tilt angle and height now, I fear the angle range available might be too small for that. Also getting the pendulum to stop at exactly the desired position might prove difficult. These all might mean that the input box might also need a motor that is synced to load the pendulum when it is in position. We shall see.

[Berthil]

Looks promising and haven’t seen this mechanism as a GBC yet, very original.

Edited September 9, 2022 by Berthil

[Heppu]

An update maybe? Or more likely a postmortem on all my attempts to make the stewart platform work as a GBC module. Unless some breaktrough happens, I'm not going to get this thing servicable for the contest.

Here is the latest version: It is too slow, unreliable and not even close to what I had in mind when starting the project. It does not capture what movements the stewart platform is truly capable. Also I use 2 extra motors at the input feeder to just make the testing easier. In the future I had plans to operate the gate that feeds 2 balls at a time to the platform by just opening it mechanically using the platform tilting down to trigger the gate.

I truly tried but no prototype really came close to what I wanted. The original pendulum idea was the one I fiddled with the furthest. The concept was 'fine' but the single axle connection meant that there was too much friction. Pendulum swings were too unpredictable and varied based on the weight of how many balls was in the scoop, often overshooting the desired ouput area with heavy loads.

I got frustrated and tried to simplify the system by making the ball track static to the platform. First I tried a weird spiral but abandoned it because it would require another module to first lift the balls to required height. (Also it blocked the view of the main mechanisms which I did not like. Then I started experimenting closer to the latest version shown in the video above with a long circular track. It would have been nice to lookat,  but the long track would have been too slow to meet the required 1ball/s requirement. That is why I decided to experiment with the more straight track shown in video instead (which is still too slow anyways). 

Anyways, good luck to anyone who is still participating! I'm kind of out of motivation to fix this machine when I know I'm most likely not going to present it at any exhibition as part of a GBC loop. I will still continue develop the stewart platform itself, but to some other purpose entirely. Cheers!

[howitzer]

Really sad to see that this didn't work out, the video shows there was a lot of promise, but it's understandable that making this work would take a lot of work. I hope that you or someone else picks up the idea and attempts to finish it as a GBC module.

[Jurss]

Don't know, but it looked like, that just needed to improve loading, and it could work.

[Berthil]

Ik really like the design and GBC potential it has, too bad it didn't work out. I also see potential in a setup where there are unexpected situations the three motor system has to react to because it is flexible/dynamic in its movements, probably more flexible than GBC needs as in GBC situation is always the same unless randomness is build in.

[Heppu]

1 hour ago, Jurss said:

Don't know, but it looked like, that just needed to improve loading, and it could work.

Sadly it's not really about the reliability but the speed. The current version processes balls at rate of ~1 ball/3sec. I would need to make it 3 times faster to be eligible for GBC standard of 1ball/sec.Big modifications would be needed to achieve that which I personally think I don't have time for atm.

1 hour ago, Berthil said:

I also see potential in a setup where there are unexpected situations the three motor system has to react to because it is flexible/dynamic in its movements, probably more flexible than GBC needs as in GBC situation is always the same unless randomness is build in.

Something like that would indeed be cool. In general I'm not sure how viable using Mindstorms in multi-person organised GBC is. If something goes wrong, it can be almost impossible to quickly fix someone elses Mindstorms project.