Holonomic drive train and field-based control

[Dan Petrovic]

Originally Posted by dmlawrence

We're considering using a holonomic drive for the upcoming season. We have a few questions for any teams that have done this in the past:

What is the approximate cost of the materials needed?
How significant is the reduction in pushing power?
What are the (dis)advantages of field-based control?
Overall, is it worth it?

Thanks,
David

Team 1751

1) it depends on where you're getting your parts.
We bought 4 56mm Banebots transmissions with encoders ~ 400 dollars.
AndyMark Mecanum Wheel set ~ 400 dollars.
You don't need to use the encoders so that would cut it down to a total of 600 dollars.

2) You wont be pushing many people, unless they have just as little traction as you. The omni/mecanum wheels have so little traction compared to today's standards.

3) Advantages: It seems more intuitive, allows for the maneuver that defines 'holonomic': Ability to rotate and translate simultaniously. Easier on the driver when robot is backwards
Disadvantages: Takes getting used to, natural bouncing of the robot due to omni/mecanum wheels may cause the gyro to do funky things

4) Drive systems will always be based on the game. In a game like last year where there is the potential for robots to be built that are unable to do anything but build until the end (ramps) I wouldn't suggest it.

[T3_1565]

we are also planning omni and with the research and designing I have done;
1. I agree with everyone else around 600-800 dollars

2. Traction is completely based on the design, I'm not allowed to go into details with this until we have completely tested the design, but I will have something up on an omni design to help with traction around build season

3. I agree with other on this too. Although it would be really nice to have a gyro for driving sake, errors with the robot getting hit, and the overall roughness of games may throw the gyro off, which in turns completely messes the driver up

4. IMHO, I have yet to see a game in which they have been a bad choice, more often than not poor proformances with omni drive systems have been based on how the driver drives the bot (trying to get into pushing matches, for example,and drving it like it was a 4 wheel drive bot). I have seen many many chances in the last few competetitions where a good (driven) omni bot could do very very well. Especially when the functional part of the bot (the arm or shooter or whatever) is designed around the drive system (for example a three axis turret for the 2006 game, so you can run around other robots by the ramp and still score points).

[David]

I agree with most of whats been said, but, my teams is also working on a prototype, we found that it will cost less than $500. We do plan on salvaging some old parts to save money, but it is enough to build a working prototype. Also I've estimated that, at best, it will have about 70% of the pushing power of a 4 wheeled robot, and, at worst, about 50%. Remember, your not trying to win pushing matches.

[Andy Baker]

Originally Posted by InfernoX14

it depends on where you're getting your parts.
AndyMark Mecanum Wheel set ~ 400 dollars.

It also depends on WHEN you're getting your parts.

We at AndyMark will be coming out with new 6" Mecanum wheels before kickoff, with a significantly decreased cost compared to the 8" version. We are shooting for an early December availability.

More details will be released soon. They will be smaller, cheaper, and with rounded (not stepped) rollers.

Andy B.

[IndySam]

Originally Posted by Andy Baker

It also depends on WHEN you're getting your parts.

We at AndyMark will be coming out with new 6" Mecanum wheels before kickoff, with a significantly decreased cost compared to the 8" version. We are shooting for an early December availability.

More details will be released soon. They will be smaller, cheaper, and with rounded (not stepped) rollers.

Andy B.

Hmmm interesting. Any other new stuff?

[JesseK]

Perhaps my idea of crab drive is a bit skewed, but with Crab drive and 4 traction wheels touching the ground, couldn't you program it to be semi-holonomic, like Team 118?

My prelim. parts list make the cost & weight pretty high compared to others (~$2k including motors, chassis $@#$@#$@#'y, and the # of victors you'll need for its control). The weight is around 50lbs depending on your machining skills (crab modules are tough to do accurately w/o machining).

For comparison, 6-wheel tank drive with 2 AM super shifters is roughly the same cost (victors included) but ~8lbs less.

[T3_1565]

holonomic and crab drive systems are different.

crab drive is a heck of a lot more complicated to make, which is why many teams vote for omni, until they have an idea on how they work, before moving on to crab drive.

Crab drive is really good IMO, but heavier and harder to make. holonomic is still very good and quite a bit easier to build. If you want JesseK look up the many other threads on the comparsions.

I think overall all these factors will be based on the game, but I would do it anyways cause I think they are really fun to watch and really fun to drive!

[Lil' Lavery]

Originally Posted by JesseK

Perhaps my idea of crab drive is a bit skewed, but with Crab drive and 4 traction wheels touching the ground, couldn't you program it to be semi-holonomic, like Team 118?

My prelim. parts list make the cost & weight pretty high compared to others (~$2k including motors, chassis $@#$@#$@#'y, and the # of victors you'll need for its control). The weight is around 50lbs depending on your machining skills (crab modules are tough to do accurately w/o machining).

For comparison, 6-wheel tank drive with 2 AM super shifters is roughly the same cost (victors included) but ~8lbs less.

Swerve (crab) systems have limitations on motion that holonomic systems do not have (but they gain full torque from their drive motors as opposed to a vector quantity, as obtained in holonomic systems). A swerve system cannot, for example, achieve accurate "frisbee motion" (spinning and translating at the same time). Additionally swerve systems cannot "instantaneously" change direction, as the wheels have to re-position alignment.
There are a great deal of different ways to build a swerve drive. The way 118 did it in 2007 didn't allow for them to change the orientation of their frame, thus requiring them to build a rotating turret for their manipulator. Contrary to this, 111's swerve allowed for traditional "tank drive" motion as well as swerve motion, but they wouldn't have been able to couple 6 drive motors together or have a multi-speed transmission nearly as easily as 118 did.
Most swerve drives do require a fair bit more machining and machine expertise than holonomic systems, as well as a fair bit more weight and money dedicated to it. But there have been lighter, cheaper, and less-complex swerve systems in the past (1261 in 2006.)

[Jeff Waegelin]

My team built a holonomic drivetrain in 2006, using DeWalt transmissions and 6" AndyMark omniwheels. The costs that people have listed above are similar to what we spent.

It was a great learning experience, and we got a lot out of the whole design process, but it turned out to be poorly suited to the game that year. We had virtually no pushing power, which made it exceedingly difficult to get into and stay in position while shooting balls into the goal with 2 defenders in the way. We also spent a lot of the season tweaking the software to improve driver control and autonomous mode - holonomic is very software intensive... without good code, your robot won't drive straight, or even at all. You also have to find the right person to drive it - we had a veteran driver, but he struggled with it the entire season. When I drove the mentor match at IRI, though, I was able to take off and go with almost no practice... some people just think the right way, and others don't.

All this being said, there are great benefits - namely the phenomenal maneuverability. Two things will help you out greatly: first, think about the design and how it fits into the game. Not all games will lend themselves well to a holonomic drive. Second, build a prototype! Unless you have tremendous resources available, and impeccable time and project management skills, you may find yourself in week 6 (or at your first regional!) still trying to get the drivetrain moving. If you do those two things, though, holonomic can be both a fun project, and a great advantage on the field.

[tseres]

here's a question for people who have done omni-directional drive systems before:
did you use sensors, and what were they? personally i think omni wheels should have a gyro, but other people think it would be better without one. also, wouldn't you need a gyro or sensors to do autonomous driving?

[Dan Petrovic]

Originally Posted by tseres

here's a question for people who have done omni-directional drive systems before:
did you use sensors, and what were they? personally i think omni wheels should have a gyro, but other people think it would be better without one. also, wouldn't you need a gyro or sensors to do autonomous driving?

We used encoders to calibrate the motors and potentially be used for autonomous. No gyro as we didn't have a field based control.

[T3_1565]

how well do you think an omni bot would drive without sensors? Like a frame omni wheels and the motors (and obviously programmed). has anyone actually tried it without encoders or gyros? (our buget it limited A LOT)

[Lil' Lavery]

Originally Posted by T3_1565

how well do you think an omni bot would drive without sensors? Like a frame omni wheels and the motors (and obviously programmed). has anyone actually tried it without encoders or gyros? (our buget it limited A LOT)

It can drive quite well, you would just have to account for any mechanical inefficiencies in your code and have a "robot-centric" control system. Just like any open-loop system, your driver would have to account for anything that goes wrong and anything that doesn't work properly.

[T3_1565]

Originally Posted by Lil' Lavery

It can drive quite well, you would just have to account for any mechanical inefficiencies in your code and have a "robot-centric" control system. Just like any open-loop system, your driver would have to account for anything that goes wrong and anything that doesn't work properly.

Ok, so its not a huge problem, unless built badly, thats good to hear thanks a lot

[rachal]

Originally Posted by dmlawrence

What is the approximate cost of the materials needed?
How significant is the reduction in pushing power?
What are the (dis)advantages of field-based control?
Overall, is it worth it?

1. Depends whether you plan to buy the wheels or make your own. My team (1072) made our own last year, the main difference between ours and AM’s being that our rollers were on the outside. We just used rubber lab stops for the rollers, which cut costs a bit. I’d say about $300 for the wheels themselves, and another $100 for the suspension (which we overengineered, I'm sure you could do it for much cheaper).

2. You get pushed by literally everything except other holonomics. Mecanums have a slight advantage over omnis in terms of traction, but if you actually want to count on pushing it’s not enough to matter.

3. Field-based control is amazing for the driver, and not so much for the programmer. We had field-oriented driving working for a while with a magnetic compass chip, but had to disable it at some point during competition. When we were testing it at school during build the robot would go crazy and bump into walls every time it passed a certain door. We eventually realized the school’s backup generator was probably hidden there

4. Heavily dependent on the game and your machining/designing skill, and in most cases no. For example, we were able to align ourselves with ramps without wasting too much time maneuvering behind the home zone line, but we got pushed around too much out in the middle of the field. So even though we could easily get in front of another bot to block, if their arm was long enough to score over us they could just shove us into the rack. Definitely not worth it for the 2007 game. However, building a holonomic is a really fun experience. Everyone should try it once, even if not for the actual competition bot

Originally Posted by tseres

here's a question for people who have done omni-directional drive systems before:
did you use sensors, and what were they? personally i think omni wheels should have a gyro, but other people think it would be better without one. also, wouldn't you need a gyro or sensors to do autonomous driving?

We just had an encoder on each wheel, and a magnetic compass for field-oriented control.

One of the downsides of mecanum is that if a wheel stops working/loses contact with the ground, the exerted forces no longer balance out, but you can still drive with two. So at some point, our programmer added code so that if a wheel went offline, its mirrored wheel would shut down.