Dynamic walking with Bioloid humanoid

[rkelbs]

Greetings All. i am investigating getting an inexpensive humanoid robot kit to use while integrating open source robot projects to allow the simulation, development, and real time control of a dynamically walking humanoid.

All control will be done on a PC to ease with debugging and development. From my limited research the Bioloid seems to be easiest setup for this over a USB connection.

I see a group at UTL has done some work (http://humanoids.dem.ist.utl.pt) but i haven't found any code or results. is anyone else working on this? any successes or failures?

has anyone had success using the "expert" bioloid kit to implement any interesting control techniques such as inverse kinematic (IK) solving with or without dynamics (center of pressure (COP)-zero moment point (ZMP))?

i'm also very interested in hearing if anyone has used the Majingaa simulator (http://www.majingaa.com/majingaa/report/v_KHR-1) with the KHR-1 and how accurate the simulations where or heard any reports.

i'm just getting started with bipeds but i am already very impressed by the accomplishments of the roboteer community and see a lot of potential for some fantastic developments!

cheers,
rob

[limor]

Hi the project at IST that you mentioned is progressing.
The page there documents how to connect Simulink to the CM5 and run a closed-loop control in real-time on the PC (by using a custom serial device driver).

The new project that is currently underway is about running the Simulink designed control loop on a gumstix. The Gumstix will be on the Bioloid connected via Jon Hyland's modules (huvrobotics.com)

unfortunately no blogged info yet.
The great outcome of the current project is that it will create an infrastructure for closed loop control of the Bioloid bus using using the Gumstix as the processing power rather than the weak Atmega.
I intend to have all the steps documented for the benefit of the Bioloid community.

[limor]

One more thing is that the EZPhysics project has been revamped in order to provide real-time physics simulation to the Bioloid. There are a couple of other physics enabled simulators. see the 3D forum threads here.

[cdraptor]

Limor

Are you using the Verdex gumstix? I bought the USB interface to the Dynamixel board that Jon created, have been sidetracked a bit playing with the Speecys SPC-101C - but I want to start doing some work on hexapod platforms - exciting news is Crustcrawler is going to release their Nomad HDATS designed for the AX-12+ so Gumstix + Jon's board + Crustcrawler HDATS would be nice combo. I've been working with the connex version since they had WiFi stick, but now the Verdex Wifi stick with miniSD is out.

Hopefully your creating some nice applcation, SDK, API or dev library that work on the Verdex and you get that in the buildroot. I'll be keeping an eye out

[rkelbs]

thanks for the knowledge! Jon's USB board combined with the gyro and pressure sensors look really nice, especially when interfaced to MATLAB/Simulink and EZphysics in real time.

i'm looking for signs that the Bioloid is capable of implementing CoM/ZMP based dynamic balancing/walking algorithms. my fear is that it is just too small, light-weight, and sensor deprived to get any useful results from. is there hope for this little guy as a research platform?

[JonHylands]

I don't see any reason why the Bioloid wouldn't make a good research robot, with the right hardware added. That's what we're trying to do with BrainBot, after all...

Of course, if I had my choice, I would ignore the AX-12 altogether, and go straight to building a custom RX-64 based robot. But figuring 20 servos at $265 each you're looking at over $5000 for just the servos, which is pricey to say the least.

If you're going to use the Bioloid, the thing we figured out pretty early is to keep the weight down as much as possible. This means using lightweight camera(s), LiPoly batteries, and such. The printed chest we're using now is fantastic for that - it weighs half what the machined Delrin chest I built weighs.

- Jon

[rkelbs]

The RX-64 would be better. Do you know if it has an improved update rate and measurements for torque than the AX-12? i don't see anything in the manual about it.

I would just use a tethered power supply for development and worry about batteries later.

[JonHylands]

I don't know. I have a couple of them here, and will be experimenting with them as soon as I mock up an RS-485 interface for them...

I'll post my results here.

- Jon

[limor]

Limor
Are you using the Verdex gumstix? I bought the USB interface to the Dynamixel board that Jon created, have been sidetracked a bit playing with the Speecys SPC-101C - but I want to start doing some work on hexapod platforms - exciting news is Crustcrawler is going to release their Nomad HDATS designed for the AX-12+ so Gumstix + Jon's board + Crustcrawler HDATS would be nice combo. I've been working with the connex version since they had WiFi stick, but now the Verdex Wifi stick with miniSD is out.
Hopefully your creating some nice applcation, SDK, API or dev library that work on the Verdex and you get that in the buildroot. I'll be keeping an eye out

Jon's solution with Gumstix works on Verdex only (USB host).
those monster crawlers with AX12 sounds promising. Maybe we'll stock them at RoboSavvy. getting it into the buildroot is a good idea (so that project "bioloid" is part of the available modules).