Close

Diagnosing the Wheel Problem

A project log for Rapidly Deployable Automation System

Unfolding CubeSat rover. Mission: construction on Mars & re-construction after natural disasters on Earth. Tele-op Headband control.

ekEK 08/27/2015 at 02:078 Comments

In the video showing the latest prototype of RDAS (v0.4), you might have wondered why I had to hold down the sides of the robot for it to move. Especially when it's supposed to be hands free -- it just seems contradictory! Time to investigate further.

Read more for explanations of the observations and the next steps.

Observations

1. Wheel not touching the ground

The wheel does not always touch the ground. If the base is not lifted high enough -- or if it is sinking in to carpet, then the edge of the side becomes in contact with the ground. This prevents the wheel from gaining direct contact with the ground, which means the torque from the spinning motor is not being applied. It isn't going anywhere because there's no friction force from the ground to overcome!

Here is what it *should* be like:

Seemingly the only time this ideal situation occurs is on the lab desk:


2. Large material deformation when force applied to wheel

When pressing on the wheel, the mount is resisting by trying to retain its original shape, which is perpendicular to the side. There is some stress in the material- the hot glue sticking the motor to the mount, and also the mount deforming. There's a large deformation along the side of the piece.

This is an issue because if the above problem was remedied, then there would still be this problem. Making the beam thicker would help alleviate it. I tried three different designs:


The one in the middle worked best.

3. The motor sides should not be able to unfold > 90 degrees

The flexible hinge that attaches the sides to the main base can move to whatever angle- both beyond the ground and also to fold up. The motors would never need to go beyond 90 degrees.

Next Steps

After diagnosing all of these issues, here are the things I need to fix:

1. Make the motor sides flush with the base so the wheels are being pushed down onto the ground

2. Move the motor down so it can extend past the edge of the side more

3. Make the motor sides only unfold to 90 degrees, BUT still be able to fold up

Time to start fixing it, v0.5 here I come


Special thanks to Reagan for looking over the diagrams

Discussions

EK wrote 09/04/2015 at 06:19 point

@Radomir Dopieralski I'm kind of excited for the linking! There doesn't have to be an electrical connection, since using mesh networking for the communication, but there will be a mechanical one. I'm really curious to see what it will be like to try to get four IR beacons aligned. (Imagining one on each corner of the side that will be linking — hard to explain without a drawing).
It's almost like you would be trying to balance two control systems at once, that are running on completely different robots. Sweet 8)

For your charging station linking, did you use any fiducials or beacons? What project was this?

Also have you seen this robot? NASA ATHLETE. http://www.nasa.gov/exploration/analogs/desertrats/desert_rats_2009_prt.htm
It reminds me of Tote a bit!

Have you ever tried to make a mechanical solution to the servos powering down problem you mentioned? Possible?

  Are you sure? yes | no

deʃhipu wrote 09/04/2015 at 09:23 point

I was trying to make one of my walking robots "sit" on a charging station in a corner, using just the distance sensor to check the distance from both walls. Perhaps that's not the best design.

There is another NASA robot that is even more similar to Tote:  (scroll to 48:00) -- It's a cheap walking robot using hobby servos!

As for a mechanical solution to the servos, I didn't try it myself, but there has been successful attempts. The cheetah cub robot uses springs to counter the gravity: 

But it can't stop at any moment -- the springs will only hold it in its standing position. The problem with this approach is that while you save energy that is needed to carry the body, once you raise a leg, you have to work against that spring, so you lose what you saved. And of course, you have to somehow accommodate the springs in the mechanical design.

  Are you sure? yes | no

deʃhipu wrote 09/04/2015 at 11:07 point

Hmm, perhaps I was a little too pessimistic in what Tote can do.

This is with the robot powered off. You can see, it can keep upright under at least 200g of tea, plus whatever the beaker weights. This is the version that uses SG92R servos for the knees, which are a little bit stronger than the cheap SG90, but I'm still quite impressed. The SG90 version collapses under that load, but can stand still without a load without a problem.

  Are you sure? yes | no

deʃhipu wrote 08/27/2015 at 13:56 point

By the way, have you considered... legs? :)

You know, something like this: 

My project is out of HaD, but I'm not giving up on it and I would still love if people used it as a base for their projects. This way you'd have a folding robot with basic mobility to start with, and could concentrate on controlling it and making it do useful stuff.

  Are you sure? yes | no

EK wrote 08/27/2015 at 16:28 point

This would be an interesting idea. I'm open to collaboration! This type of locomotion would be decent on an exploring module.

The whole idea of the RDAS platform is that multiple robots with different purposes can be created— then linked together to perform tasks.


One thing that concerns me is carrying the weight of the bot on those servos. Do you have a link to a project update where you discuss this? Think I saw a vid where Tote was carrying a box, or similar.

I'll make a bare RDAS chassis and weigh it, will let you know how much it weighs (this might take a day or so).

Also were you considering of a different way of attaching the servos so it is not directly attached? Springs, or some type of suspension, to not harm the servos when there is an impact.

  Are you sure? yes | no

deʃhipu wrote 08/27/2015 at 17:10 point

I don't want  yet another project, but I would be glad to help with any issues you have with #Tote's code or design. I assume you would 3D-print your own mechanics anyways, but I'm happy to share my experiences in what works there and what doesn't, and help you adjust the code for the new hardware.

I'm not sure how easy linking would be. I have attempted to make some of those robots link to a charging station, and found that it's quite a difficult task. I'm concerned that doing it with two robots that are both moving at the same time may be even harder. Not that legs are any worse than wheels here -- they can even be better, as you can move sideways.

I don't really have a log about carrying weight (But I should have it, thank you for the idea!). The heaviest stuff I had #Tote carry was a box of 3 AA alkaline batteries in https://hackaday.io/project/6050-tote-affordable-spider-robot/log/18738-different-power-sources . As you can see, it manages surprisingly well, but becomes sensitive to center of mass not being at the center of the robot. And those servos are working at ½ of their proper voltage. If SG90 servos are not strong enough, you can swap them with SG92R, which are even stronger.

As for attaching the servos, those plastic servo horns that I use are actually quite flexible, which lowers the robot's accuracy the ability to tell how far it moved from the initial position, but provide protection to those servos. I had this robot fall from my desk to the floor without damage.

The real drawback of legs is actually the power consumption. Those robots can walk around continuously in a circle for 20 minutes on a single 200mAh battery charge. With wheels you would probably get an hour of driving around. What's worse, while the gait is statically stable and you can just stop at any moment without falling, unpowering the servos will make the robot "sit down" on the floor. Doing that to recharge the battery would be possible, but could result in disproportionate time spent not moving.

  Are you sure? yes | no

deʃhipu wrote 08/27/2015 at 06:58 point

Hmm, that seems to be a serious problem. Even if you make the wheels stick out more (Wouldn't that be a problem when the robot is folded? It's no longer a cube on the outside!), the base will always just lie there on the ground. Seems like you either need a locking mechanism, or some springs that would keep the flaps open and lower than the base.

  Are you sure? yes | no

EK wrote 08/27/2015 at 13:48 point

Agreed! Making the wheels stick out more would be a problem for the bot to fit into a CubeSat deployer. (Unless the deployer is modified to allow more room on two of the sides, that would be nice) I talk about an idea for a mechanism in the latest update:

https://hackaday.io/project/6647-rapidly-deployable-automation-system/log/24165-wheel-improvements-with-4-elastic-bands


The base does have casters (marbles) on it; so at least it is only 4 points of relatively small contact compared to the entire square of the base.



See the latest update for how the sides are kept open, plus the new rigid hinges.

  Are you sure? yes | no