Proposal for an Alternate Parallax MadeUSA® Base Design
Hello,
I had an idea and I thought this might be the best place to get feedback. I originally put this in the Blogs area, but davejames kindly suggested that I move it here for visibility. I originally wrote this paper a while ago so it might be dated, but I think it still applies. I will attach the Word document but also add it to the post in case some readers don't have Word.
The idea has to do with a simple re-arrangement o the MadeUSA / Eddie / Arlo platforms to solve a problem in my home environment. Specifically, uneven terrain.
I am anxious to hear what you think.
dtaylor (Doug)
===========
I had an idea and I thought this might be the best place to get feedback. I originally put this in the Blogs area, but davejames kindly suggested that I move it here for visibility. I originally wrote this paper a while ago so it might be dated, but I think it still applies. I will attach the Word document but also add it to the post in case some readers don't have Word.
The idea has to do with a simple re-arrangement o the MadeUSA / Eddie / Arlo platforms to solve a problem in my home environment. Specifically, uneven terrain.
I am anxious to hear what you think.
dtaylor (Doug)
===========
Comments
dtaylor (Doug)
Numerous books cover the topic of chassis design well. One of my faves is Braunl: http://www.amazon.com/Embedded-Robotics-Mobile-Applications-Systems/dp/3540034366 which discusses the merits of each, among many other topics.
A simple way to interpret your redesign is to make the existing base smaller (cutting off the rear), put 2 casters up front and shift the weight more forward. That solves some problems and changes others. I agree with you that people tend to gravitate toward circular, zero-radius chasses, but whether those are actually better for a "smart bot" that really shouldn't be bumping into walls & corners remains to be seen. IMO, sensors and software are good enough and cheap enough to mostly keep the bot away from entanglements. A lot comes down to personal preference and even what's available. I suspect there are a lot of bots used in university and lab studies that mostly deal with flat, open concrete or linoleum spaces. They may not have to deal with thresholds, uneven floors, or the odd sweater or magazine laying on the floor. And stairs are right out...
From my experience, the best way to get a point across is demonstrate an advantage. If you have built something, please feel free to show it here. I have built a few bots, and no matter how simple they are, I learn something new each time. Here's a bot I built several years back, similar to your proposal except with a single caster. It was amazingly accurate at dead reckoning, even on carpet and transitioning to & from a tile floor: http://www.youtube.com/watch?v=PX0IhUqnwrk
I read your proposal and I think that this amazing lecture from David Anderson focus on the problems that you describe.
You can also you try a Styngray's Robot design.
Have you seen the video of David Anderson's presentation? I learned a lot from it. He just about convinced me of the advantages of a round robot with wheels on the outside edge. I thought his point "if a robot can get stuck it will get stuck," one which was interesting to think about.
I personally would like my house robot to be able to climb stairs.
I second erco's suggestion of showing us your robot if you have one built.
It might be easier for others to read your paper if you convert it to a pdf.
Edit: I see Nikos bet me to recommending David's video. Good to see I'm not the only one who liked it.
Duane,
To say the truth I saw for the fist time David's video following one of your links!
Thanks for your suggestions! One other thing that we'll throw into the re-design mix is to add shocks to the caster assemblies. Haven't really had time to pencil it out just yet, although our desire is to deliver the very best, robust, and cost effective robotic platforms on the market.
As erco suggested, a great way to test concepts is to build/modify different variations of what we're all thinking. It's really cool to see how different builders tackle different challenges:-)
We'll keep you posted on our end - we're a pretty "open-discussion" type of company, and in fact many of our products are the way they are because of customer feedback...thanks!
-MattG
I enjoyed reading your thoughtful analysis of these design considerations. Indeed, it's an interesting topic since one design is never ideal for every situation. Prior to the base we have today, there were some prototypes that were more reflective of your ideas. Here's a really early picture from before we even got into the CNC routing process, made from a handheld jig saw:
As you pointed out, this design is incredibly effective at navigating floor height transitions. It pays a minimal penalty as a three-wheeler by being able to do an "almost ZTR" rotation. You mentioned that the programming issues are relatively simple to add, and I concur. But the sensing challenges can be really big, and good programming requires good sensing. Consider a "land mine" of an office with rolling chairs, legged tables, black baseboards and people who wear pants. Detecting a metal chair leg is really difficult since it's very narrow; people's pants absorb sound waves from the ultrasonic sensor, and black baseboards absorb infrared light. Add to these challenges a robot rear end that can get caught up during a turn and navigation becomes more complicated.
The design featured last in your Word document is a good compromise. Drive wheel placement supports ZTR and tail-dragging wheels are in the back so there's no high-centering. Getting creative with the battery placement could avoid having the robot tip forward during a sudden stop. About the only drawback [assuming you can design out the forward-tip by weight placement] is how cool it looks [or doesn't look]. And since beauty of robots is personal and totally subjective, it doesn't matter what anybody else thinks except the builder of the robot.
Could we make something like you propose? You bet. It's a question of when we could fit it in, whether or not it's a small prototype run, and whether or not there's compatibility with our existing vacuum table fixtures. This last point is a big one, as productive CNC routing is all about investing in proper fixture design. You see, the fixtures sit between the vacuum table and the piece of material being machined. The fixtures have holes so the air suction can hold down the work. If we can do alternate designs without violating the work with new holes everything is quite easy. You'd almost have to see the fixture to know what I mean. Again, not a big deal but a consideration.
Here's the original thread where we first discussed making these bases (caution, these first efforts look like a hack!):
http://forums.parallax.com/showthread.php/118838-Robotics-Base-for-the-12V-Motor-Mount-Wheel-Kit-(27971)-and-Caster-Wheel-(28971
Since then we've moved the manufacturing process into Parallax with a far more productive Haas CNC router, complete with tool changer, vacuum table, dust collection, etc.
I really enjoy this discussion and encourage it to continue. If you want me to send you some HDPE blanks to cut them up the way you want I'd be happy to do that. This stuff cuts like butter and takes drills easily, so you could plunge a jig saw into a piece of material and make it any way you want. Let me know if you want some material - I'd be happy to support you in this effort.
Carry on with the great discussion. It'll probably be a matter of minutes before Gordon McComb weighs in. He knows a bunch about this topic.
Ken Gracey
I took a minute to re-read Erco's post above. He's a resident genius with this topic and has lots to share (especially about where to find good parts for cheap). Somehow we're able to get his free advice all day long. Nikos has a lot of experience with this exact platform and drive system, and Duane works with all of the products we offer. And MattG makes it happen inside Parallax. Without him....well, don't want to think about that possibility!
Ken Gracey
Not totally free, Ken. You generous Parallaxians (you, Matt, Jim, and more) shower us all with great products through contests, giveaways, mystery bags & prizes, so it's in our best interest to give back and keep the ball rolling.
I am overwhelmed by all the response. Thank you all for the positive feedback. I'm going to go back and read each in detail and try to respond to everyone, but this silly day job keeps getting in the way.
Keep the ideas coming.
dtaylor (Doug)
Thanks for the reply.
---
"it's likely you have reached your conclusions based on doing some real-world experimentation" - Yes, I built a small robot with the round body and forward and backward casters and immediately ran into the uneven terrain problem.I have also built a couple of tricycle designs such as the BoeBot and some BoeBot hacks and a quadcycle design with a similar robot called Trekker from SuperDroid. The design I put forth for the Arlo is more of a thought experiment at this point, but I'm starting to get fired up about it.
http://www.superdroidrobots.com/shop/item.aspx?itemid=25
I don't think they sell the kit anymore, but apparently they have remaining stock of the chassis.
---
I checked out the book link on Amazon and they specify It is a hardcover, but only give a paperback price. I might look around for it. Thanks for the pointer.
---
The concept of putting the castors in front is interesting. Clearly you have done this with your Retrobot Pathfinder. Great bot by the way. This brings up an interesting question I hadn't thought of. Would there be any advantage/disadvantage to having the castors pushed or pulled?
I suppose you can drive your Retrobot Pathfinder either direction, have you designed any experiments to test this? Is it easier to push a castor over an obstacle such as a thick rug or to pull it over the same obstacle?
---
You mentioned the Retrobot Pathfinder is amazingly accurate at dead reckoning. As I understand it, you drove the robot and recorded the data that took it back to the starting position and then played it back and was off only by an inch. That is quite an accomplishment. However, when you were driving the robot over the uneven surfaces, the resulting data took into account the terrain variances. The playback would be adjusted to the small twists and turns produced by the terrain.
Could you program in the data simply using geometry and get the same level of response? Just curious.
Thanks,
dtaylor (Doug)
Ok, I still can't paste. Can anyone tell me what I'm doing wrong?
dtaylor (Doug)
This is text that was <cntl> c’d and <cntl> v’d from word into this forum post.
All I did was type the text in Word, highlighted the text that I wanted to copy, held the CTRL key down and the pressed the letter "c". Then I placed my cursor into the forum window, and then held the CTRL key down and then pressed the letter "v".
is that working for you?
-MattG
99.9% of the members can read a PDF.
Jim
Thanks Jimbo!
-MattG
Matt,
I thought it was my work computer, but I can't paste from my home computer either. I type into Notepad, select the text, <cntl> c then select the edit window and <cntl> v and nothing happens. I know it is in my paste buffer because I can paste back into Notepad. Don't know what I'm doing wrong.
Jim,
Thanks for the helping hand.
we're missing something here...hummm - which browser you using?
C'mon you Forumista BrainTrusta-farians...get Jim fixed up :thumb:
-MattG
The now-discontinued Stingray used this arrangement, with one of those funky omniwheels. I wonder if a single one of those, heavy-duty enough, would be a practical alternative. Arlo is too big to use the simple roller ball concept of the BOE Bot.
In talking about the front/back caster design, I think it's still important to keep in mind that the Arlo/Eddie/MadeUSA is designed just like the typical lab robot, which is made for the ordinary classroom, which has low-nap carpet or tile. Though the Parallax design does not use it, it is fairly common to set the casters so they are at different heights, and then place the bulk of the weight on the longer one. The front is then constantly doing a wheelie. The robot can teeter over obstacles like carpet thresholds. This effectively is the same as a 3-point base, but with the zero turn radius, added weight capacity, and simplified coding for the students.
Matt,
you can't read the PDF?
I'm using Chrome.
Jim
Let me ask though, have you done any odometry with a robot in that configuration? A little off axis turn might be more than you thought. It might be wise to try it and see how you like it before you decide that it is more or less capable.
I thought that Parallax MadeUSA® Base Design was designed mainly for classrooms. I would imagine that its odometry is as good as it gets. I think that anyone (especially myself) who designs a robot will try to encompass a lot of design features to make it as multipurpose as possible. However, if I were a student, in a classroom, learning and coding a robot I think that the current platform is great.
If there were a second base design, where do you stop?
First you get it to go over a one inch bump, but then there's a two inch bump you need to go over. Then a three inch bump and before you know it you'll need it to climb stairs. So in reality the need for something different will always be there.
I think that you can keep the current design and switch out the casters for something spring loaded as mentioned earlier. My old Roomba Discovery had a nice spring loaded caster-like wheel and it was perfect for a much smaller robot. I'm sure there are other options available.
As great as your design is, I would look at different casters, softer, with a little suspension before doing anything else.
-MattG
Matt and Jim,
I seem to have created some confusion. Matt, the post you were responding to was from me (dtaylor (Doug)). In that same post, I addressed Jim and thanked him for his contribution. You saw "Jim" and thought it was a signature. Sorry.
Now, to answer Matt's question - I am using IE 11 in both places. I will try Chrome a little later.
Matt,
I really like the casters with shock absorbers idea. In the process of design, think about a mechanism to stiffen the movement such as allowing the user to insert washers to compress the spring. Something simple like that. I look forward to what you come up with.
Thanks for the great response. I have been a fan of Parallax for years.
Your prototype tricycle with the single tail caster looks like it should handle my thresholds and carpet transitions perfectly. Also, it is quite attractive. What is the clearance with that design. I know the weight of the batteries should be below the plane of the wheel's axes for stability, but if the batteries are too thick they will compromise the ground clearance. I just measured my worst threshold and it is one and one sixteenth inch high.
Doug - we're looking forward to what you come up with too :thumb:
-MattG
Matt,
Ok, I just tried it with Chrome. When I have something in my paste buffer and I then click the "reply" link and get the "quick reply" edit window. Then I click the "paste as plain text" button, then it pops up a window in which I can paste the text. It all works fine in Chrome.
In fact this is pasted in while using Chrome.
In IE 11 I never get the popup paste window so it looks like I can paste into the main edit window, which I can't.
I just checked IE 11 and I do NOT have popups blocked.
Fair point Xanadu, This could be just my solution to my problem. It was originally just a thought experiment anyway.
Nikos G. and Duane,
I just finished watching the whole video and it was fantastic. Thanks for the pointer.