How are teams mounting drivetrain encoders on WCD?

[gc_coxen]

Originally Posted by AdamHeard

The drive axle is closer in gearing to the floor than somewhere near the motor, it will have a much more accurate measurement of floor position.

My mistake, I meant to say between mounting to the drive axle or a separate wheel axle.

[Lil' Lavery]

I don't have any pictures unfortunately, but a couple of 1712s students designed a novel mount for the Versa block system. They 3D printed a piece that interfaced with the contour of the Versa block above the 2x1 channel (including the existing mounting holes) and stick out just beyond the end of our drive axle. There was an adjustable encoder mount that was then slid into that overhung lever, allowing a structural element to hold the encoder that could also slide with the Versa block as the block moved to tension chain.

[marshall]

Originally Posted by Lil' Lavery

I don't have any pictures unfortunately, but a couple of 1712s students designed a novel mount for the Versa block system. They 3D printed a piece that interfaced with the contour of the Versa block above the 2x1 channel (including the existing mounting holes) and stick out just beyond the end of our drive axle. There was an adjustable encoder mount that was then slid into that overhung lever, allowing a structural element to hold the encoder that could also slide with the Versa block as the block moved to tension chain.

Would be cool to get a picture of this. Sounds like a cool add-on.

[Oblarg]

Originally Posted by Lil' Lavery

I don't have any pictures unfortunately, but a couple of 1712s students designed a novel mount for the Versa block system. They 3D printed a piece that interfaced with the contour of the Versa block above the 2x1 channel (including the existing mounting holes) and stick out just beyond the end of our drive axle. There was an adjustable encoder mount that was then slid into that overhung lever, allowing a structural element to hold the encoder that could also slide with the Versa block as the block moved to tension chain.

This is an exceedingly smart idea that I think we will have to steal next time we do a DS drive.

[GKrotkov]

Originally Posted by Lil' Lavery

I don't have any pictures unfortunately

Here we go

[Lil' Lavery]

A couple quick notes on those pictures.


The one prominanetly displayed there was actually used with a VersaBlock we had on a manipulator joint, rather than our drivetrain. We didn't intend to have that VersaBlock slide, thus why we fastened the encoder mount down to the 2x1 aluminum with the bolt you can see. One of the drivetrain encoder mounts can be seen in the background of a couple of the pictures. For the drivetrain mount, you can see the bolts that are used for the VersaBlock are also used to secure the encoder mount.



The electrical tape was a pit repair since a crack developed in the piece that holds the encoder itself. Next year we'll likely beef up that piece. Unfortunately the tape obscures the design somewhat, but the idea is that the piece that holds the encoder can slide on the main mount structure in order to find your ideal fit. If you know your shaft lengths beforehand, you could make this all one print custom fitted to work with your shaft overhangs.

[Oblarg]

I absolutely love that versablock encoder mount idea, and will probably urge 449 to steal it

[AdamHeard]

Originally Posted by gc_coxen

My mistake, I meant to say between mounting to the drive axle or a separate wheel axle.

FWIW, 254 generally (or always?) mounts their encoder on a wheel shaft and has some of the fanciest autos out there.

[Nuttyman54]

Originally Posted by AdamHeard

FWIW, 254 generally (or always?) mounts their encoder on a wheel shaft and has some of the fanciest autos out there.

Just to make sure people understand here: Adam's example is valid proof that putting an encoder on an outer wheel axle will not inherently add too much variability to produce reliable autos.

I feel that it's worth adding that this is possible because 254 properly tensions their chains, which means the backlash/slop between the outer wheels and directly driven wheels is minimal. If you're using WCP sliding versablocks, you should be able to tension your chain/belt well enough to make this a viable solution as well.

As an aside, 254 and 971 gave a great talk at Champs 2015 which is recorded on Youtube about Motion Profiling and Control which is super helpful to understand how they get such reliable auto control. Worth looking at.

TLDR; If you have a reasonable tensioning method for your chains (esp sliding bearing blocks) you should be fine putting an encoder on one of the outer wheel axles.

[AdamHeard]

Originally Posted by Nuttyman54

Just to make sure people understand here: Adam's example is valid proof that putting an encoder on an outer wheel axle will not inherently add too much variability to produce reliable autos.

I feel that it's worth adding that this is possible because 254 properly tensions their chains, which means the backlash/slop between the outer wheels and directly driven wheels is minimal. If you're using WCP sliding versablocks, you should be able to tension your chain/belt well enough to make this a viable solution as well.

As an aside, 254 and 971 gave a great talk at Champs 2015 which is recorded on Youtube about Motion Profiling and Control which is super helpful to understand how they get such reliable auto control. Worth looking at.

TLDR; If you have a reasonable tensioning method for your chains (esp sliding bearing blocks) you should be fine putting an encoder on one of the outer wheel axles.

I don't believe 254 ran tensioners in 2016 or 2015.

[Nuttyman54]

Originally Posted by AdamHeard

I don't believe 254 ran tensioners in 2016 or 2015.

Correct. I actually had a much longer post drafted up and then deleted most of it to keep it shorter and more straightforward. Tension your chains and you shouldn't have a problem.

I would consider a well-designed exact C-C setup to be "properly tensioned", such as 254 ran in 2015 and 2016 and 118 for the past few years as well.

Some quick math also shows that a few degrees of "slop" of the wheels relative to eachother will only result in a positional error of fractions of an inch (1 degree of slop on a 4" diameter wheel is less than 0.040" linear). While you may get some "noise" in your encoder values, you should be able to tune a PID loop to be stable with that amount of error and still give extremely precise auton navigation.

[Thad House]

Originally Posted by Nuttyman54

Just to make sure people understand here: Adam's example is valid proof that putting an encoder on an outer wheel axle will not inherently add too much variability to produce reliable autos.

I feel that it's worth adding that this is possible because 254 properly tensions their chains, which means the backlash/slop between the outer wheels and directly driven wheels is minimal. If you're using WCP sliding versablocks, you should be able to tension your chain/belt well enough to make this a viable solution as well.

As an aside, 254 and 971 gave a great talk at Champs 2015 which is recorded on Youtube about Motion Profiling and Control which is super helpful to understand how they get such reliable auto control. Worth looking at.

TLDR; If you have a reasonable tensioning method for your chains (esp sliding bearing blocks) you should be fine putting an encoder on one of the outer wheel axles.

I think think another trick is making sure you put the encoder on the weighted wheels. That way you don't have to worry about chain tension, and the encoder should always be touching the ground, at least in most cases. Another thing some teams do is preload their chains. Instead of pushing your robot forward into its starting position, push it too far forward, and then pull back. This will preload the chains to they are tensioned in the right direction for moving forward, which helps reduce shock loads.

[Chris is me]

Originally Posted by Thad House

I think think another trick is making sure you put the encoder on the weighted wheels. That way you don't have to worry about chain tension, and the encoder should always be touching the ground, at least in most cases. Another thing some teams do is preload their chains. Instead of pushing your robot forward into its starting position, push it too far forward, and then pull back. This will preload the chains to they are tensioned in the right direction for moving forward, which helps reduce shock loads.

Isn't this backwards? If the robot last moved backwards, when forward voltage is applied it will have to drive through all of the slop in the system before moving forward. If the robot last moved forward, the slop is "behind" the applied power to the system, and it won't jerk as much. So you want the last "push" of a robot being set up for auton to be forward.

It's a good idea to push the robot back and forth a few times as you set it up, in either case, to make sure everything is running smooth and your drivetrain isn't stuck on a burr or anything like that. I've just always made sure the last push was in the direction the robot intends to travel.

[Thad House]

Originally Posted by Chris is me

Isn't this backwards? If the robot last moved backwards, when forward voltage is applied it will have to drive through all of the slop in the system before moving forward. If the robot last moved forward, the slop is "behind" the applied power to the system, and it won't jerk as much. So you want the last "push" of a robot being set up for auton to be forward.

Its definitely pull if you want to go forward. If your motor is to the left of the wheel, when the motor attempts to drive the wheel, it will pull the bottom length of the chain tight. Pulling the wheel backwards also causes the bottom part of the chain to go tight, which is what you need.

[AdamHeard]

Originally Posted by Chris is me

Isn't this backwards? If the robot last moved backwards, when forward voltage is applied it will have to drive through all of the slop in the system before moving forward. If the robot last moved forward, the slop is "behind" the applied power to the system, and it won't jerk as much. So you want the last "push" of a robot being set up for auton to be forward.

It's a good idea to push the robot back and forth a few times as you set it up, in either case, to make sure everything is running smooth and your drivetrain isn't stuck on a burr or anything like that. I've just always made sure the last push was in the direction the robot intends to travel.

You're thinking in the wrong reference frame, forward in the motor frame is reverse when externally pushed.