Darwin-LC(not OP) CAD files
Darwin-LC(not OP) CAD files
by wiweet
Mon Oct 15, 2012 3:30 pm
From various source of information, I've found that on the development/prototyping phase Darwin-OP is called Darwin-LC (Low cost),Darwin-LC uses AX-18 servo. my question is, where could I found more Info (detailed specs, CAD files, etc) for the Darwin-LC? is it open to pubilc as Darwin-OP does?
Because due to my budget, I want to build the Darwin-LC, not the Darwin-OP.
Thanks
Because due to my budget, I want to build the Darwin-LC, not the Darwin-OP.
Thanks
by Gort
Mon Oct 15, 2012 11:03 pm
wiweet, I have seen this rumor too and possible pictures of a prototype OP with AX-18s.
I have not seen the plans for this version yet and I don't think you will see them from a official source.
I have also heard the rumor that this prototype was not a very good robot. The servos just did not have enough torque for the weight. I would agree with this because of my experiences with my bioloid humanoid which has AX-18s in his legs. The speed is great but the holding torque is not much better than AX-12s.
Even with that issue i think a version of the OP with AX-18 is a interesting idea and I will be my next project after I finish building my OP clone. my idea is to modify the existing brackets for the AX-18 and find ways to save weight. The AX-18 OP may not have covers because of the extra weight they add to the robot. I also think the best way to save weight is to 3d print out all of the brackets.
I am also not sure how much money you well really save with this version?
I am thinking only about 1 to 2 thousand USD? My OP clone has cost me about 6000 USD to make. So I think you will spend about 4000 USD to build a AX-18 version of the OP.
I have not seen the plans for this version yet and I don't think you will see them from a official source.
I have also heard the rumor that this prototype was not a very good robot. The servos just did not have enough torque for the weight. I would agree with this because of my experiences with my bioloid humanoid which has AX-18s in his legs. The speed is great but the holding torque is not much better than AX-12s.
Even with that issue i think a version of the OP with AX-18 is a interesting idea and I will be my next project after I finish building my OP clone. my idea is to modify the existing brackets for the AX-18 and find ways to save weight. The AX-18 OP may not have covers because of the extra weight they add to the robot. I also think the best way to save weight is to 3d print out all of the brackets.
I am also not sure how much money you well really save with this version?
I am thinking only about 1 to 2 thousand USD? My OP clone has cost me about 6000 USD to make. So I think you will spend about 4000 USD to build a AX-18 version of the OP.
by wiweet
Tue Oct 16, 2012 7:22 am
Thanks for the quick reply Gort.
Honestly, my plan is not exactly build an exact Darwin-LC, particularly in the controller part.
My idea begins from my previous project, when I build robot ( 2 robot) to participate in my country Robocup league. My design was based on Bioloid Premium robot, DMP Vortex86 SBC and a webcam. I have succesfully build a vision system on the Vortex86 SBC, that issued a motion command to the CM-510 controller. The major problem is that the servo I use (AX-12) is not have proper torque to handle the extra weight coming from the Vortex86 SBC and webcam.
So I think an upgrade using AX-18 would improve my design, and I came across Darwin-LC video that uses AX-18, it makes me want to build a Darwin-LC, at least has the same kinematic design with it (Darwin-LC).
I also have another alternative to my next project, it basically start from a bioloid premium, but I will use the RX-24F in the 3 lowest part of the leg(heel,toe,knee) and use AX-18 and AX-12 at the upper part.
Talking about my budget, I will have around $1600 for each robot just to improve the kinematic, not the controller and the electronics.
Is it a viable idea and budget? or you have another suggestion for me?
Honestly, my plan is not exactly build an exact Darwin-LC, particularly in the controller part.
My idea begins from my previous project, when I build robot ( 2 robot) to participate in my country Robocup league. My design was based on Bioloid Premium robot, DMP Vortex86 SBC and a webcam. I have succesfully build a vision system on the Vortex86 SBC, that issued a motion command to the CM-510 controller. The major problem is that the servo I use (AX-12) is not have proper torque to handle the extra weight coming from the Vortex86 SBC and webcam.
So I think an upgrade using AX-18 would improve my design, and I came across Darwin-LC video that uses AX-18, it makes me want to build a Darwin-LC, at least has the same kinematic design with it (Darwin-LC).
I also have another alternative to my next project, it basically start from a bioloid premium, but I will use the RX-24F in the 3 lowest part of the leg(heel,toe,knee) and use AX-18 and AX-12 at the upper part.
Talking about my budget, I will have around $1600 for each robot just to improve the kinematic, not the controller and the electronics.
Is it a viable idea and budget? or you have another suggestion for me?
by PedroR
Thu Oct 18, 2012 11:23 am
Hi all
I just wanted to jump in and add some contribution about DARwIn-LC.
To be honest this thread was the first we heard of the "LC" and Googling for it seems to bring up interesting results:
http://techcrunch.com/2010/06/28/meet-d ... oid-video/
http://www.youtube.com/watch?v=amiUhtSrHLI
I must say it looks a lot like a Bioloid GP with a DARwIn head.
I have posted in the past that the GP looked a lot like an early prototype of DARwIn and having seen the pics in the links above, I am becoming more and more confident that it was probably the case.
The metal work is quite similar to DARwIn (especially the upper torso and the place where the boards are placed in DARwIn (that space is currently empty in the GP)).
Between the GP and DARwIn there is an important difference though which is the leg design.
I'm not 100% sure of the differences but I know the servos are not exactly in the same configuration.
Regarding the torque of the AX servos, I wouldn't be able to say much, but I know i-Bot built a Robot running the DARwIn framework (gait generator, vision, etc) using AX servos and Beaglebone.
As I understand, to make your job easier, the only two things you should ensure is that the Leg servos are mechanically in the same position and that you use the same IMU (same chips).
You can then configure the dimensions of your Robot in the Framework code to have the gait generator working almost out of the box.
I'm surprised no one's ever considered taking a GP and just building the extra Metal frames needed to make it into a DARwIn configuration.
Cost-wise you'd end up with something similar to Darwin (full metal frame robot but with AX servos) at a fraction of the price (probably ~$5k)
Regards
Pedro.
I just wanted to jump in and add some contribution about DARwIn-LC.
To be honest this thread was the first we heard of the "LC" and Googling for it seems to bring up interesting results:
http://techcrunch.com/2010/06/28/meet-d ... oid-video/
http://www.youtube.com/watch?v=amiUhtSrHLI
I must say it looks a lot like a Bioloid GP with a DARwIn head.
I have posted in the past that the GP looked a lot like an early prototype of DARwIn and having seen the pics in the links above, I am becoming more and more confident that it was probably the case.
The metal work is quite similar to DARwIn (especially the upper torso and the place where the boards are placed in DARwIn (that space is currently empty in the GP)).
Between the GP and DARwIn there is an important difference though which is the leg design.
I'm not 100% sure of the differences but I know the servos are not exactly in the same configuration.
Regarding the torque of the AX servos, I wouldn't be able to say much, but I know i-Bot built a Robot running the DARwIn framework (gait generator, vision, etc) using AX servos and Beaglebone.
As I understand, to make your job easier, the only two things you should ensure is that the Leg servos are mechanically in the same position and that you use the same IMU (same chips).
You can then configure the dimensions of your Robot in the Framework code to have the gait generator working almost out of the box.
I'm surprised no one's ever considered taking a GP and just building the extra Metal frames needed to make it into a DARwIn configuration.
Cost-wise you'd end up with something similar to Darwin (full metal frame robot but with AX servos) at a fraction of the price (probably ~$5k)
Regards
Pedro.
by PedroR
Thu Oct 18, 2012 11:38 am
There's an interesting detail in the youtube link I posted above.
If you notice the back exposes a grid in metal that's probably designed for ventilation. If you've ever used a DARwIn, you know that grid doesn't exist and in it's been upgraded to 2 active fans that vary speed depending on whether they're on wall power or battery power (and they can get noisy I must add).
It's also interesting to see how Darwin "shrunk down" looks much cutter than the actual DarwIn (ie same, design but changing scale changes our perception and feel about the Robot).
Regards
Pedro.
If you notice the back exposes a grid in metal that's probably designed for ventilation. If you've ever used a DARwIn, you know that grid doesn't exist and in it's been upgraded to 2 active fans that vary speed depending on whether they're on wall power or battery power (and they can get noisy I must add).
It's also interesting to see how Darwin "shrunk down" looks much cutter than the actual DarwIn (ie same, design but changing scale changes our perception and feel about the Robot).
Regards
Pedro.
by Gort
Thu Oct 18, 2012 6:48 pm
Pedro,
Yes I saw those pictures and the video.
A AX-18 DARwIn-Op is my next project after my clone OP is done. I think a 3D printed version would be lighter and a better robot then the all metal LC. It sounds like the LC was over weight and top heavy. 3D printed parts may solve this issue? The Raspberry PI is also looking like a good replacement for the FitPC too.
Yes I saw those pictures and the video.
A AX-18 DARwIn-Op is my next project after my clone OP is done. I think a 3D printed version would be lighter and a better robot then the all metal LC. It sounds like the LC was over weight and top heavy. 3D printed parts may solve this issue? The Raspberry PI is also looking like a good replacement for the FitPC too.
by PedroR
Thu Oct 18, 2012 7:13 pm
Hi Gort
I would actually love to see the results you're getting with the 3D Printed frames.
From the experiments we've made, it seems it is possible to 3D print the frames but you need to do see work on the CAD file to reinforce key structural zones. (see for example the post about the Robot Limor's designing).
Regarding the Metal vs Plastic it is possible a "Darwinized" bioloid GP was overweight.
From our experience, we've built a Bioloid GP with 18 servos (torso rotation and head rotation) and we found it was the weight of the extra servos that created some difficulties. (a stock GP has only 16 servos; a Darwin has 20)
From what we've seen the most sensitive part is the Ankle joint and the knee. (this is inline with what you said).
I guess you can work around these issues by putting an MX-28 on the lower body (it uses the same AX bus but the commands have higher resolution; there are work arounds for this).
Regarding the raspberry PI, the DArwin framework comes as source code.
Therefore, as long as you have a Linux OS, you should be able to compile the framework to run on any host.
The gait generator uses a lot of FPU but I trust a Raspberry PI would be able to handle that.
Going back to iBot's project he used a Beaglebone which I think is also ARM (I just don't know how the Bone and the PI compare).
There is an intermediate step which i-Bot did but I don't remember how he achieved it, where I think he has communication flowing from the PC to the servos though an intermediate board (like the CM-730 does).
Using the intermediate board is a very nice idea because you can rewrite the commands if need to.
(for example if the Darwin framework is sending commands with 4096 resolution, the intermediate board can rewrite them to 1024 of the AX (or let it pass through for any MX servos you may have in the mix).
Finally, if you're looking to use a PI, I assume you'd be looking to replace the CM-730. I recommend you have a look at iBot's project.
he's used an Arduino (a Pro mini I think), added the same IMu chips (he got them from Pololu I think) and program it to emulate the same memory mapping of the CM-730 et voilá.
(there's a lot of work before the voilá
but he's been able to prove the concept really well!)
Anyway, I apologise for the long reply.
I wrote a lot more than a reply to your question. I guess I'm just excited about the idea and getting carried away
Your approach of going with the DARwIn electronics kit is probably the most clever t this stage as you can now focus on fabricating the mechanics.
We get a lot of inquiries from people looking to get the CM-730 only (probably for custom projects where they want to run the DARwIn framework), but Robotis won't sell it separately. Yet interestingly no one has achieved a drop in CM-730 clone/replacement afaik.
Regards
Pedro
I would actually love to see the results you're getting with the 3D Printed frames.
From the experiments we've made, it seems it is possible to 3D print the frames but you need to do see work on the CAD file to reinforce key structural zones. (see for example the post about the Robot Limor's designing).
Regarding the Metal vs Plastic it is possible a "Darwinized" bioloid GP was overweight.
From our experience, we've built a Bioloid GP with 18 servos (torso rotation and head rotation) and we found it was the weight of the extra servos that created some difficulties. (a stock GP has only 16 servos; a Darwin has 20)
From what we've seen the most sensitive part is the Ankle joint and the knee. (this is inline with what you said).
I guess you can work around these issues by putting an MX-28 on the lower body (it uses the same AX bus but the commands have higher resolution; there are work arounds for this).
Regarding the raspberry PI, the DArwin framework comes as source code.
Therefore, as long as you have a Linux OS, you should be able to compile the framework to run on any host.
The gait generator uses a lot of FPU but I trust a Raspberry PI would be able to handle that.
Going back to iBot's project he used a Beaglebone which I think is also ARM (I just don't know how the Bone and the PI compare).
There is an intermediate step which i-Bot did but I don't remember how he achieved it, where I think he has communication flowing from the PC to the servos though an intermediate board (like the CM-730 does).
Using the intermediate board is a very nice idea because you can rewrite the commands if need to.
(for example if the Darwin framework is sending commands with 4096 resolution, the intermediate board can rewrite them to 1024 of the AX (or let it pass through for any MX servos you may have in the mix).
Finally, if you're looking to use a PI, I assume you'd be looking to replace the CM-730. I recommend you have a look at iBot's project.
he's used an Arduino (a Pro mini I think), added the same IMu chips (he got them from Pololu I think) and program it to emulate the same memory mapping of the CM-730 et voilá.
(there's a lot of work before the voilá
but he's been able to prove the concept really well!)
Anyway, I apologise for the long reply.
I wrote a lot more than a reply to your question. I guess I'm just excited about the idea and getting carried away
Your approach of going with the DARwIn electronics kit is probably the most clever t this stage as you can now focus on fabricating the mechanics.
We get a lot of inquiries from people looking to get the CM-730 only (probably for custom projects where they want to run the DARwIn framework), but Robotis won't sell it separately. Yet interestingly no one has achieved a drop in CM-730 clone/replacement afaik.
Regards
Pedro
by wiweet
Fri Oct 19, 2012 9:01 am
A paper related to Darwin LC coul be found here
http://www.ais.uni-bonn.de/humanoidsoccer/ws10/papers/HSR10_02.pdf
http://www.ais.uni-bonn.de/humanoidsoccer/ws10/papers/HSR10_02.pdf
by i-Bot
Fri Oct 19, 2012 4:50 pm
In my experience if you want to achieve a really low cost Darwin, then you must keep the weight down. The AX18 does not appear to give a major torque improvement, so if you get too heavy and break the weight barrier(probably around 2kgs), you may as well go MX28 and associated $$.
On my DARwin-pi, I used micro servos for the head pan tilt to cut weight. The Raspberry Pi is much lower weight than Fit-PC, but has some performance limitations. The CM-730 is useless for a low cost/low weight DARwin, it is only really used as a communications controller, and performance is usually better if it is left out. I still retain the CM510 in the DARwin-pi, but only as for power control/monitor/distribution, this means low cost upgrade from premium. Later upgrade will be to replace CM510 with smaller power control and hopefully second battery.
All done with standard frame parts except Raspberry pi mounting which is laser cut acrylic. Only 3D printed part will be new head when I get around to it. Weight is now just over 1.9kgs including pan tilt camera and all electronics.
Result is both of my DARwin-pis cost less than $1000 complete since I bought premiums on eBay, but even with a new premium they would have cost less than $1500 each. They run a modified version of the Robotis DARwin software. The UPenn software is likely more of a challenge on the pi.
On my DARwin-pi, I used micro servos for the head pan tilt to cut weight. The Raspberry Pi is much lower weight than Fit-PC, but has some performance limitations. The CM-730 is useless for a low cost/low weight DARwin, it is only really used as a communications controller, and performance is usually better if it is left out. I still retain the CM510 in the DARwin-pi, but only as for power control/monitor/distribution, this means low cost upgrade from premium. Later upgrade will be to replace CM510 with smaller power control and hopefully second battery.
All done with standard frame parts except Raspberry pi mounting which is laser cut acrylic. Only 3D printed part will be new head when I get around to it. Weight is now just over 1.9kgs including pan tilt camera and all electronics.
Result is both of my DARwin-pis cost less than $1000 complete since I bought premiums on eBay, but even with a new premium they would have cost less than $1500 each. They run a modified version of the Robotis DARwin software. The UPenn software is likely more of a challenge on the pi.
by wiweet
Fri Oct 19, 2012 6:05 pm
Thanks for your advice i-Bot. About your Darwin-pi, have you post it at robosavvy's forum?if you already did, could you please point me to it?so I could have more insight to your Darwin-pi.I am also interested in using micro servo for the head pan and tilt, in fact I already have 3 metal gear micro servo in my part inventory.