PDA

View Full Version : [Question(s)] Arbotix or CM-510 Controller?



Jaybots
02-22-2011, 03:40 PM
I'm about to take the plunge and build a Bioloid humanoid robot. I have decided to go with ax-12s servos to get my feet wet. My question for the moment is which controller--the Arbotix or CM-510. I hope to chase that illusive goal of the "perfect humanoid walk" among other things.. I have a reasonable amount of software background and wired my share of boards, so I am more the type who likes to roll up there sleeves and get right in there.

Given that little introduction, which controller are people using for their humanoids and why.
-----------------------
PS. I have been lurking on this board for a while an want to thank all the contributors, it has been very useful.

DresnerRobotics
02-22-2011, 04:12 PM
I'd say this totally depends if you plan on implementing IK or not.

From a purely FK pose based standpoint, the CM-510 is going to have better software for you to create motions. Roboplus is basically built to do exactly this. With that said, it's pretty limited in logic based software and communications options.

The ArbotiX is going to be a more flexible option, however it's best suited for development of an IK based walking gait. I do believe that lnxfergy started work on this and might very well be willing to share. You'll have more IO options and a solid communication solution this route too.

Ultimately, an IK based walking gait generator is a much better way to go, but has considerably more work involved in getting it up and running. FK pose based walks are more instant-gratification, however are SEVERELY limited in the long run. This is speaking from experience using both controllers to build humanoids.

iBot
02-23-2011, 06:45 AM
You should start with the CM510. This controller and the associated software and motions are the essential get you started kit fordeveloping bioloid humanoids. Even if you then develop your own motions, they will most likely be based heavily on the CM5/510 experience.

I would think the premium kit with CM510 is similar price to buying parts without anyway.

The Arbotix is OK if you want Xbee to a remote PC, sort of a wireless USB2 Dynamixel.

With the Darwin developments, you will see new controller options this year, which are more appropriate to the whole range of Bioloid humanoids with IMU and associated IK software.

Jaybots
02-23-2011, 09:48 AM
Thanks for your responses. I tend to agree with you Tyberius that a true natural walk can only be achieved through IK coupled with dynamic feedback. This of course leads to the next question, that is, if autonomous natural walking is the objective , are either processors up to the challenge. Does anyone have experience using a floating point coprocessor (Micromega's uM-FPY for example) to handle the heavy number crunching with either of these two processors.

iBot, I agree that purchasing the parts separately gets you almost to the cost of Bioloid premium and that I would learn a lot from such a setup, but what am still struggling with is not so much of the dollar cost of getting the Bioloid premium first and then getting the Arbotix as my needs grow, but the investment in time in first learning all the mechanics of the CM-510 in order to get as much out of it as possible only to have to start over again with the Arbotix.

iBot, the Darwin developments you were talking about, are they the work being done by RoMeLa and if so all I could find is this http://www.romela.org/main/DARwIn_OP:_Open_Platform_Humanoid_Robot_for_Resear ch_and_Education . I also found a block diagram of the robot showing a new CM-730 as a sub controller and main controller as SBC-fitPC2i. Do you have any other links?

DresnerRobotics
02-23-2011, 11:08 AM
The Arbotix is OK if you want Xbee to a remote PC, sort of a wireless USB2 Dynamixel.



While I agree that the CM-510 is easier for beginners and has my recommendation to people looking to do humanoids with FK based gaits, the Arbotix is far more than a wireless USB2Dynamixel (which has zero onboard smarts/IO). The application of it in context of humanoids would use the onboard processing for IK and interpolation, and only use the wireless for high level commands.

iBot
02-23-2011, 01:34 PM
I would suggest neither the CM510 or the Arbotix processor is up to the task of IK based gaits alone. Their value is in their ability to demonstrate proven FK ability and to act as a sub controller in an IK configuration. Though the Arbotix, still seems to offer little advantage in either role unless you are building a Mech. Adding external FPU will not resolve this performance deficit, you need a full bore main controller with integrated FPU for the IK and dynamic gait state machines.
btw. Don't undervalue FK + human crafted moves! or Uncle Bob will have you in the dust.

To understand the Darwin resources take a look at the UPenn source for an amazing view of not just IK. but also IMU integration and even ZMP and vision.
http://sourceforge.net/projects/darwinop/files/Software/Main%20Controller/Source%20Code/DARwIn-OP_UPenn_Open_Source.zip/download
Of course this uses a FitPC, but much of the processing power is for vision, but do think at least linux based processor with FPU such as Gumstix or similar OMAP device.

To see the Robotis perspective, look at
http://darwin-op.springnote.com/pages/6861909

One thing to note is how little demand is made of the CM-730 sub controller in both the UPenn and the Robotis code for the motion part.

lnxfergy
02-23-2011, 01:42 PM
I would suggest neither the CM510 or the Arbotix processor is up to the task of IK based gaits alone. Their value is in their ability to demonstrate proven FK ability and to act as a sub controller in an IK configuration. Though the Arbotix, still seems to offer little advantage in either role unless you are building a Mech.

I've tried to stay out of this thread, but I have to call shenanigans on that claim. You certainly can't do high-dimension search based (RRT, RDT, etc) IK on a 16Mhz processor -- but you can do an optimized, forward-equation based approach to IK. In fact, I have done 4/5/6 DOF Biped IK -- at approximately 40-50hz on the ArbotiX, with space to spare. Yes, there is still work to do in integrating an IMU, but certainly, it can be done.

As for the ArbotiX on a biped, there hasn't been a huge amount of work in that particular area -- most of our development has been in quad/hexapod IK.

-Fergs

Disclaimer: I created the ArbotiX, so I may be a bit biased...

Slugman
02-23-2011, 05:34 PM
I am probably only a few months further along than Jaybots with regard to robotics, so from a Noob perspective, could you experts please don't forget what he is asking. I.E. You have his stated rough goal, so what should he get that will allow him to learn how to use one thing, then be able to apply that knowledge to further develop/upgrade his robot. Very technical arguments are useful for people like Fergs & Tybs, 'cos that's what they do, but for people like me (& I'm guessing Jaybots is included in that), that's not what he needs to know.
I myself have gone for the Arbotix over the CM-xxx purely because I wanted to incorporate my own programming in an RC robot controlled from a PC, & I felt that even though it would be a steeper learning curve, it would be of benefit in the long run, as even if I 'upgraded' to a Fit-PC or some-such later, the programs I had already put together would still work but I could use the greater processing power for on-board vision processing etc etc

The question is not whether either of those options CAN do it, but which one has the greater developmental capacity FOR HIS STATED PURPOSE with his level of expertise. (I'm not yelling, just emphasising) To put it another way - He is going to buy a motherboard for a computer. Which motherboard will allow him to attach the best sound/video cards or RAM etc so he can get the best life out of the motherboard before he has to upgrade.

I took a slight leap of faith when I got the Arbotix, as I have limited programming experience, but apart from my stupid questions on this forum, I made it though to make a walking quad (Now Hex & am working on the claws), so hopefully he can do the same, given the very helpful nature of the guys here giving us all their hard work as open source software. Just my 2c worth. :happy:

tician
02-23-2011, 11:53 PM
I may be a bit late to the discussion, but from my experiences with the CM-5/510/700 I cannot recommend them. Too many caveats in both hardware and firmware: You can wirelessly upload task files to a CM-5, but not to a CM-510/700. You cannot use RoboPlus Motion or Manager wirelessly. (both of the previous points are not due to hardware limitations, just firmware/software that Robotis does not presently intend to address (http://www.robotis.com/xe/qna_en/28883)) The ZIG-110 for the CM-510/700 just dangles on a wire by the connector and must be secured by plastic rivets. The CM-700 has only one 4-pin connector to share between a ZIG-110 for wireless communication and an LN-101 for programming and motion editing (i.e. constant (un)plugging during debug and cannot reprogram a ZIG-110 except through task files). No external connectors/breakouts on the CM-510 except with the recessed 5-pin headers for which they do not sell spare cables (http://www.robotis.com/xe/?mid=qna_en&page=2&document_srl=35698).

That said, I have never even touched an arbotix. One possible disadvantage of the arbotix is the use of an ATmege644 whereas the CM-510 uses an ATmega2560 (mostly just more memory). To me, the arbotix still seems like the better choice with the readily accessible GPIO/PWM headers (hobby servos or anything else), the onboard motor driver, the onboard XBee connector and transparent wireless control (please correct me if I am wrong on this point, but this feature alone makes me covet the arbotix), the ROS driver, NUKE and Pypose, and finally because of Fergs himself for his frighteningly active presence on the forums.



Wow. I really love to rant against Robotis, don't I? Sorry about that. They do make really great servos though.

lnxfergy
02-24-2011, 12:23 AM
I think this all really boils down to what type of person you are. The CM-X models are targeted at a different audience than the ArbotiX -- neither will make everyone happy.

The Robotis CM-X models and their associated software are generally targeted at people who are non-programmers. They offer nice drag and drop modules that make it a snap to get basic software on the bot. They also only support Windows (as far as I know).

The ArbotiX is targeted at programmers -- who just happen to want to use dynamixel servos. We offer a basic programing suite, including all the bells and whistles (and tutorials and documentation) of the Arduino IDE, and our custom libraries for using dynamixels, onboard motor drivers, etc. The real power of the ArbotiX is the ability to interface just about anything else you want to it (extra sensors, airsoft guns, etc). ArbotiX-based bots have won in Mech Warfare, at the Trinity Fire Fighting Home Robot contest, and a number of events at the Canadian National Robot Games. If you do happen to go towards a PC-based bot later on (using, say, a FitPC), the ArbotiX's support for sync_read (which avoids many of the latency issues caused by USB) and ROS are quite nice features.

-Fergs

iBot
02-24-2011, 06:14 AM
No "shenanigans" intended from me. I guess a topic titled " Arbotix or CM-510 Controller?" was bound to raise some interesting comment.

If the question is simply " Arbotix or CM-510 Controller?" , then I agree with the above comparison from Fergs, though I would also add this comment from Tician,
and finally because of Fergs himself for his frighteningly active presence on the forums
Maybe Fergs is too modest to stress the importance of the support currently available for the Arbotix and Arduino. In this context I endorse the choice of controller by Slugman for his needs,and have also endured the frustrations of tician with the CM510

My replies were based on the whole question from Jaybots including the subtext, especially:

build a Bioloid humanoid robot and

perfect humanoid walk

This sets the bar a lot higher if the intent is to walk better than existing Bioloid humanoid. My proposal is still to initially pursue development with the readily available walks and gaits of the CM510 and follow the work from Uncle Bob to improve the stock walk motions.
http://robosavvy.com/forum/viewtopic.php?t=6556

Why not Arbotix for this? this is also the answer to:

which controller are people using for their humanoids and why.

There has been little support for humanoids on the Arbotix to date.


The subsequent question from Jaybots

This of course leads to the next question, that is, if autonomous natural walking is the objective , are either processors up to the challenge
This is the where to go from there question, and prompted some of the more technical alternatives on FK, IK, IMU, dynamic walking etc.
In this solution space neither the CM510 or the Arbotix has established presence, being mainly dominated by onboard linux devices. Maybe Tyberius and Fergs can change that ? Giger upgrade to Arbotix maybe.

Jaybots
02-24-2011, 09:04 AM
Wow, what a great set of comments, and lively too. Thanks a lot. I also think Fergs nailed it for me in describing the two different type of users . I know I am going to real bored real fast watching the pre programmed moves in the CM-510, clap, flex, etc. So I think for me it boils down to access and future potential. My quick look at the specs of the two processors suggest that they are roughly comparable with the 644p having a hardware multiplier and 2561 having more memory so the hardware itself is not a factor.

I suspect that initially to achieve what I after I am going to have a pc do all/most of the calculations and then work on moving that down to the bot level. In this case communication speed becomes paramount. So I think score one for the arbotix. I am currently using Arduinos and processing, so that only makes the arbotix more attractive. Tician's description of what the CM-510 can not do seal the deal for the arbotix

The question about whether any Atmega processor is up to the calculation requirements remains. In looking around at other people's projects I could not help but notice Tyberius' last two projects used Gumstix processors (I also noticed the incredible craftsmanship and attention to detail on your bots). I suspect the Gumstix choice may be more for the build-in video and communication capabilities? But what would be the next level up processor that people are using?

Update: Sometimes fate intervenes to resolve the issue. I found a really good deal on a Bioloid Premium kit which I could not pass up, so I took the plunge. Given the deal, I am also going to get arbotix and will probably play around first with the cm-510 and then move over to the arbotix. Now, while I wait for the postman, I can work on a mathematical model of a Type A Bioloid-- anyone have any weight and dimension data on the various segments?

tician
02-24-2011, 07:32 PM
Congrats on the purchase. If we really are Sith Lords (http://forums.trossenrobotics.com/showthread.php?p=41440#post41440) for using Dynamixels, then welcome to the Dark Side.
I wish you the best of luck with the CM-510 and RoboPlus, they really are great systems but they can become inadequate very quickly if you have much ambition and/or programming experience. Also, hopefully you will not have the crashing problems I have had with RoboPlus (probably due to using Windows XP and clicking way too quickly while trying to program in RoboPlus Task).

As to the AX-12+ and its brackets, there are CAD files for most if not all of the brackets in the Data Center (http://forums.trossenrobotics.com/datacenter/robotis-33/bioloid-ax-12-brackets-151/) and the dimensional drawings (http://support.robotis.com/en/techsupport_eng.htm#product/dynamixel/ax_series/dxl_ax_actuator.htm) and weight for the AX-12+ are on the Robotis Support site. The bracket weights may be listed on Robotis' online shop or they may possibly be considered negligible (at least compared to the AX-12+ and LiPoly battery of the Premium kit).

DresnerRobotics
02-26-2011, 09:52 PM
No "shenanigans" intended from me. I guess a topic titled " Arbotix or CM-510 Controller?" was bound to raise some interesting comment.


I actually agree entirely with your assessment of a CM-510 being better for a humanoid/beginner user. I was just pointing out that it's a bit unfair/inaccurate to refer to the Arbotix as merely a wireless USB2Dynamixel.