darwin motions
by Fritzoid
Wed Jun 01, 2011 2:44 pm
Hi limor,
I uploaded a file called BioSim.zip in my files area. It contains the updated Darwin-OP source code I used to generate walking-gait position data. I modified the walk tuner for the Bioloid kinematics and then captured the sync_write commands so that I can feed them into the simulator.
The real number crunching is done in module Walking.cpp. It does a lot of floating point operations, mostly trigonometric functions and some square roots. It's probably too much to do on a RISC processor but not too much for a board with an FPU. As far as ZMP goes the Darwin code doesn't seem to use it. However it can be calculated in the OpenHRP simulator.
My plan is to take the position data that runs on the simulator and to create a motion file from it. I have a program that allows me to chart the motions graphically and then to select key frames which are converted into motion file steps. While it's not real-time control it does allow a poor Bioloid user to get some use out of the Darwin framework. Once I get the download function working I should have something I can share.
I uploaded a file called BioSim.zip in my files area. It contains the updated Darwin-OP source code I used to generate walking-gait position data. I modified the walk tuner for the Bioloid kinematics and then captured the sync_write commands so that I can feed them into the simulator.
I'm wondering just how much processing power is required to drive the inverse dynamics and ZMP.
The real number crunching is done in module Walking.cpp. It does a lot of floating point operations, mostly trigonometric functions and some square roots. It's probably too much to do on a RISC processor but not too much for a board with an FPU. As far as ZMP goes the Darwin code doesn't seem to use it. However it can be calculated in the OpenHRP simulator.
My plan is to take the position data that runs on the simulator and to create a motion file from it. I have a program that allows me to chart the motions graphically and then to select key frames which are converted into motion file steps. While it's not real-time control it does allow a poor Bioloid user to get some use out of the Darwin framework. Once I get the download function working I should have something I can share.
by Fritzoid
Wed Jun 08, 2011 3:12 pm
Like most long running threads this one has digressed a bit from its original intent. Mostly we have been looking to modify the Darwin code so that it can be used by a Bioloid.
BioSim.zip is a package that contains the customized files that I used to simulate the Bioloid Premium using OpenHRP. Part of this project involved modifying the Darwin code to reflect the differences between a Darwin and a Bioloid. You can modify the Darwin framework modified with the changes in BioSim.zip to use the Darwin walk tuner with a Bioloid. The code can be used to drive real servos over a USB2Dynamixel link or stored as a position file for the OpenHRP simulator. You can find the package at http://robosavvy.com/Builders/Fritzoid/
BioSim.zip is a package that contains the customized files that I used to simulate the Bioloid Premium using OpenHRP. Part of this project involved modifying the Darwin code to reflect the differences between a Darwin and a Bioloid. You can modify the Darwin framework modified with the changes in BioSim.zip to use the Darwin walk tuner with a Bioloid. The code can be used to drive real servos over a USB2Dynamixel link or stored as a position file for the OpenHRP simulator. You can find the package at http://robosavvy.com/Builders/Fritzoid/
by robaldo
Wed Jun 08, 2011 8:59 pm
Fritzoid wrote:Like most long running threads this one has digressed a bit from its original intent. Mostly we have been looking to modify the Darwin code so that it can be used by a Bioloid.
BioSim.zip is a package that contains the customized files that I used to simulate the Bioloid Premium using OpenHRP. Part of this project involved modifying the Darwin code to reflect the differences between a Darwin and a Bioloid. You can modify the Darwin framework modified with the changes in BioSim.zip to use the Darwin walk tuner with a Bioloid. The code can be used to drive real servos over a USB2Dynamixel link or stored as a position file for the OpenHRP simulator. You can find the package at http://robosavvy.com/Builders/Fritzoid/
hi
thanks for your attention
which files should i download?
just http://robosavvy.com/Builders/Fritzoid/BioSim.zip?
also please tell me about OpenHRP! which link should i download for windows? there is some links in download page!
by Fritzoid
Thu Jun 09, 2011 12:36 pm
Hi robaldo,
If you want to use the Darwin-OP code you are going to need a Linux machine. You will also find that the OpenHRP simulator is a lot easier to install and use in Linux than in Windows. Now if you happen to have a PC running Ubuntu 10.04, you're all set. If not consider doing what I did and install it as an application on your Windows system. Then you can choose a boot time which system to run. I do pretty much all my work in Linux these days.
As far a software, you will need the Darwin-OP framework which can be downloaded from SourceForge at http://sourceforge.net/projects/darwinop/files/ Download and extract DARwIn-OP_ROBOTIS_vxxx.zip.
The source code will then need to be edited to include the changes found in my BioSim.zip package. This is because my changes were made to version 1.0.0 and the latest version is 1.2.0 (or later). The readme.txt file lists the source code files that have changed and the updates are commented in each file.
If you want to use the simulator it will be necessary to install OpenHRP. This turns out to be a much bigger deal than any other software you've probably ever installed. The instructions on how to do this can be found at http://www.openrtp.jp/openhrp3/en/install.html
I have installed OpenHRP and have it running on Windows XP but it took me a couple of days. I was unable to get it working at all on my Windows 7 box because it (Windows) runs in 64-bit mode. So unless your a glutton for punishment go the Linux route.
If you want to use the Darwin-OP code you are going to need a Linux machine. You will also find that the OpenHRP simulator is a lot easier to install and use in Linux than in Windows. Now if you happen to have a PC running Ubuntu 10.04, you're all set. If not consider doing what I did and install it as an application on your Windows system. Then you can choose a boot time which system to run. I do pretty much all my work in Linux these days.
As far a software, you will need the Darwin-OP framework which can be downloaded from SourceForge at http://sourceforge.net/projects/darwinop/files/ Download and extract DARwIn-OP_ROBOTIS_vxxx.zip.
The source code will then need to be edited to include the changes found in my BioSim.zip package. This is because my changes were made to version 1.0.0 and the latest version is 1.2.0 (or later). The readme.txt file lists the source code files that have changed and the updates are commented in each file.
If you want to use the simulator it will be necessary to install OpenHRP. This turns out to be a much bigger deal than any other software you've probably ever installed. The instructions on how to do this can be found at http://www.openrtp.jp/openhrp3/en/install.html
I have installed OpenHRP and have it running on Windows XP but it took me a couple of days. I was unable to get it working at all on my Windows 7 box because it (Windows) runs in 64-bit mode. So unless your a glutton for punishment go the Linux route.
by robaldo
Thu Jun 16, 2011 7:19 am
thanks
i will download the OpenHRP and use the simulation under Ubuntu 11.04 I've installed.
but my main question is that how could i use the code with ax-12s?
what i have is listed
a CM-510
a USB2Dynamixel
a RB-110
i think there is no need of CM-510.
so i want to use RB-110 and USB2Dynamixel.
in the DARwin-OP code there are lots of files.
i wonder which ones to use? (just for walk)
i also wonder if you help me make the humanoid robot walk with ax-12 actuators.
thanks
i will download the OpenHRP and use the simulation under Ubuntu 11.04 I've installed.
but my main question is that how could i use the code with ax-12s?
what i have is listed
a CM-510
a USB2Dynamixel
a RB-110
i think there is no need of CM-510.
so i want to use RB-110 and USB2Dynamixel.
in the DARwin-OP code there are lots of files.
i wonder which ones to use? (just for walk)
i also wonder if you help me make the humanoid robot walk with ax-12 actuators.
thanks
by i-Bot
Thu Jun 16, 2011 10:57 am
For the hardware you only require the RB-110. The USB2Dynamixel is already integral to the RB-110. The CM510 is not needed either.
Since you have no CM-730, you need to do the modifications to the CM730 files I describe here:
http://robosavvy.com/site/index.php?opt ... 7&Itemid=2
These were based on Darwin 1.0 and if you use subsequent, you will have to look into any issues raised by the 1024/4096 switch and the support of sync read. Also if you want to add some emulation of CM-730 features such as buttons, leds, Gyro, etc. based on RB-110 peripherals you should do it there also.
This should get you set up to use Roboplus to your DARwIn RB-110 with ax-12. Only use the roboplus linux project in DARwIn, the others have embedded servo positions which ard unlikely to be compatible with your robot. Use Roboplus motion files for your robot, not ones for Darwin-OP. Best to test first with servos outside of robot, or adapt simulator for your ax-12 robot.
If you use dxl_monitor, avoid the reset command because it sets DARwIn-OP specific values.
The next stage in using the DARwIn code outside Roboplus depends on the robot configuration (DOF, kinematics, camera) you have built with the ax-12. Can you describe this so I can help describe the files to modify?
Since you have no CM-730, you need to do the modifications to the CM730 files I describe here:
http://robosavvy.com/site/index.php?opt ... 7&Itemid=2
These were based on Darwin 1.0 and if you use subsequent, you will have to look into any issues raised by the 1024/4096 switch and the support of sync read. Also if you want to add some emulation of CM-730 features such as buttons, leds, Gyro, etc. based on RB-110 peripherals you should do it there also.
This should get you set up to use Roboplus to your DARwIn RB-110 with ax-12. Only use the roboplus linux project in DARwIn, the others have embedded servo positions which ard unlikely to be compatible with your robot. Use Roboplus motion files for your robot, not ones for Darwin-OP. Best to test first with servos outside of robot, or adapt simulator for your ax-12 robot.
If you use dxl_monitor, avoid the reset command because it sets DARwIn-OP specific values.
The next stage in using the DARwIn code outside Roboplus depends on the robot configuration (DOF, kinematics, camera) you have built with the ax-12. Can you describe this so I can help describe the files to modify?
by robaldo
Thu Jun 16, 2011 12:46 pm
as i understood, you mean i should use RoboPlus and i don't know why should i use RoboPlus while i want to use dynamic walk of DARwin? (i mean the code)
let me describe more...
i have a Bioloid premium (type A) with ax-12 actuators.
i want to use DARwin's walk file with ax-12.
so what do i need in order to do that?
could you please tell me step by step!?
I am confused because plenty of information you give me!
could you please describe me, what every folder in DARwin-op is used for?
please behave me like a beginner!
thanks
let me describe more...
i have a Bioloid premium (type A) with ax-12 actuators.
i want to use DARwin's walk file with ax-12.
so what do i need in order to do that?
could you please tell me step by step!?
I am confused because plenty of information you give me!
could you please describe me, what every folder in DARwin-op is used for?
please behave me like a beginner!
thanks
by Fritzoid
Fri Jun 17, 2011 12:47 pm
Hi robaldo,
The updates to the darwin framework that I posted (BioSim.zip) include both my and i-Bot's modifications.
i-Bot's update allows the software to pretend that it's has a built-in controller. It is enabled by setting EMULATE_CM730 to true in the CM730 constructor. The idea is that by running the software on a PC you can do away with the SBC-fitPC2i part of the darwin system. Then by emulating the CM-730 you can do away with the need for a controller too. That leaves the dynamixels as the only actual hardware and they can be driven over a serial port using a USB2Dynamixel. Thus is detailed in i-Bot's wiki posting.
My update goes further and allows you to pretend that you have a string of dynamixels as well. You enable my code by setting EMULATE_RX28M true also. Then instead of sending commands to the servos the software will save the commands to a file. The resulting file can then be used to run a simulation or it can be imported into motion file using my Motion Edit program.
My updates also include the modifications to the kinematics engine that are necessary to run the walk_tuner with a Bioloid type A. So to use the walk tuner with a real type A it would be necessary to apply my changes, set ENABLE_CM730 true but to leave ENABLE_RX28M false. Then you need to rebuild the project in ../darwin/Linux/project/walk_tuner. Running the resulting program will then drive the servos from a Linux machine.
At this point a note of caution would seem prudent. THIS IS A VERY RISKY OPERATION!!! Your first attempts will NOT work correctly. The robot can easily be damaged from a tumble or colliding parts. You need to build some type of harness ahead of time and be prepared to kill the power fast!
As i-Bot mentioned, the modifications were designed for version 1.0.0 of the darwin framework. The current release is version 1.2.0 so the code has changed in substantial ways. I have been working with the new code for just a day now and already I've had to make numerous changes just to get it to compile. I will post the updates once I have the code working properly.
The updates to the darwin framework that I posted (BioSim.zip) include both my and i-Bot's modifications.
i-Bot's update allows the software to pretend that it's has a built-in controller. It is enabled by setting EMULATE_CM730 to true in the CM730 constructor. The idea is that by running the software on a PC you can do away with the SBC-fitPC2i part of the darwin system. Then by emulating the CM-730 you can do away with the need for a controller too. That leaves the dynamixels as the only actual hardware and they can be driven over a serial port using a USB2Dynamixel. Thus is detailed in i-Bot's wiki posting.
My update goes further and allows you to pretend that you have a string of dynamixels as well. You enable my code by setting EMULATE_RX28M true also. Then instead of sending commands to the servos the software will save the commands to a file. The resulting file can then be used to run a simulation or it can be imported into motion file using my Motion Edit program.
My updates also include the modifications to the kinematics engine that are necessary to run the walk_tuner with a Bioloid type A. So to use the walk tuner with a real type A it would be necessary to apply my changes, set ENABLE_CM730 true but to leave ENABLE_RX28M false. Then you need to rebuild the project in ../darwin/Linux/project/walk_tuner. Running the resulting program will then drive the servos from a Linux machine.
At this point a note of caution would seem prudent. THIS IS A VERY RISKY OPERATION!!! Your first attempts will NOT work correctly. The robot can easily be damaged from a tumble or colliding parts. You need to build some type of harness ahead of time and be prepared to kill the power fast!
As i-Bot mentioned, the modifications were designed for version 1.0.0 of the darwin framework. The current release is version 1.2.0 so the code has changed in substantial ways. I have been working with the new code for just a day now and already I've had to make numerous changes just to get it to compile. I will post the updates once I have the code working properly.
by i-Bot
Sat Jul 16, 2011 1:40 pm
I put up a new draft article on DARwin OP here:
http://robosavvy.com/site/index.php?opt ... 1&Itemid=2
It covers the joint layouts and compares to other humanoids with a view to using the DARwin OP software framework from Robotis on other humanoids.
I now have the DARwin software framework working on Robobuilder. This required creation of support for the wCK servo. Also substancial changes were required to support the playing of Robobuilder .rbm files and to support the slightly different joint layout and kinematics. No gyro or accelerometer yet.
Still tuning the walking routines, but I will put up another article when I get some time.
http://robosavvy.com/site/index.php?opt ... 1&Itemid=2
It covers the joint layouts and compares to other humanoids with a view to using the DARwin OP software framework from Robotis on other humanoids.
I now have the DARwin software framework working on Robobuilder. This required creation of support for the wCK servo. Also substancial changes were required to support the playing of Robobuilder .rbm files and to support the slightly different joint layout and kinematics. No gyro or accelerometer yet.
Still tuning the walking routines, but I will put up another article when I get some time.
by i-Bot
Tue Jul 19, 2011 12:28 pm
I never thought to check before, but it appears ASIMO has also the same leg configuration as DARwin with 6 DOF per leg. Lots more DOF in ASIMO arms though.
http://world.honda.com/ASIMO/technology/spec.html
http://world.honda.com/ASIMO/technology/spec.html
