Planning to build HR-OS1 robot using AX-12A & AX-18A (looking for frames)

[kshin]

Hi, I am considering taking AX-12A's from my PhantomX/Hexapod to build a HR-OS1. I also have a handful of AX-18A, so I would probably use it as well. I was wondering if there is anyone still actively pursing HR-OS1? Also, I was wondering if I still can buy the Frames from Trossen Robotics. I saw the kit on the website, but was wondering if it the inventory is still available. Thanks!

Kevin

[tician]

I'm not aware of what happened internally at Trossen, but I've not seen much of several of their long-time employees and their development platforms have shifted almost exclusively to ROS and robot arms/rovers.

Humanoids are very difficult in terms of both hardware costs and software development, but the HROS-1/5 were intended to decrease the hardware costs through a fairly inexpensive, open-source, standardized design using relatively few custom parts compared to many humanoid development systems that are either one-off experiments in research labs or require long-term company support via closed-source software for the hardware to do anything useful. Unfortunately, everything was initially built on providing compatibility with the DARwIn-OP framework which is nice for getting started quickly (and playing soccer) but was rather difficult to expand/experiment. It seems that adoption of the hardware was not very widespread compared to the stock Robotis Bioloid kits that can be slightly reconfigured and augmented with additional servos and a new control board to give a bot very similar in design to the HROS-1.

All of the metal frames are still available individually in the Trossen store, but the HROS specific parts do not appear to be available in the shop any longer. They might have some leftovers hanging around the place that they could sell if you email them directly. Models for the required 3D printed and laser-cut parts can be extracted from the models repository (https://github.com/Interbotix/HROS1-3D-Models). If you are really desperately interested in a true HROS-1, I have two printed head/neck tilt printed part kits and at least some of the laser cut parts and other kit components (http://forums.trossenrobotics.com/sh...off-some-stuff). Using laser cut acrylic and plastic standoffs for the HROS-1 torso never sat well with me, so intended to replace them with 3D printer torso box to hold all the electronics in a stronger, stiffer, safer design. Unfortunately, I backed a 3D printer on kickstarter instead of buying a kit...

Overall, humanoids are still a bit too specialized compared to wheeled rovers with arms, or even just non-humanoid bipeds. All that extra weight swinging around high off the ground makes control schemes an area of active research while not providing much benefit to non-researchers. A humanoid torso on a wheeled base is much easier for interactions as you don't have to deal with weight, balance, and power/runtime issues anywhere near as severe as with a legged humanoid. A arm/torso-free biped is often adequate for research into bipedal gaits with cost savings on the arms that can go into better servos for the legs.

One of my biggest regrets is that I received a free HROS-1 from Tyberius but never actually did much with it. If only it had gone to a researcher/educator actually capable of being productive instead of me. Ripley has been in a partly assembled state for nearly three years without so much as a minute of actual software development towards making her walk again. I've been trying to make myself partly reassemble her to make a MechWarfare biped (legs with swivel/tilt torso holding airsoft turret), but there are lots of things I've been trying to make myself do for years while failing miserably.

[kshin]

I'm not aware of what happened internally at Trossen, but I've not seen much of several of their long-time employees and their development platforms have shifted almost exclusively to ROS and robot arms/rovers.

Humanoids are very difficult in terms of both hardware costs and software development, but the HROS-1/5 were intended to decrease the hardware costs through a fairly inexpensive, open-source, standardized design using relatively few custom parts compared to many humanoid development systems that are either one-off experiments in research labs or require long-term company support via closed-source software for the hardware to do anything useful. Unfortunately, everything was initially built on providing compatibility with the DARwIn-OP framework which is nice for getting started quickly (and playing soccer) but was rather difficult to expand/experiment. It seems that adoption of the hardware was not very widespread compared to the stock Robotis Bioloid kits that can be slightly reconfigured and augmented with additional servos and a new control board to give a bot very similar in design to the HROS-1.

All of the metal frames are still available individually in the Trossen store, but the HROS specific parts do not appear to be available in the shop any longer. They might have some leftovers hanging around the place that they could sell if you email them directly. Models for the required 3D printed and laser-cut parts can be extracted from the models repository (https://github.com/Interbotix/HROS1-3D-Models). If you are really desperately interested in a true HROS-1, I have two printed head/neck tilt printed part kits and at least some of the laser cut parts and other kit components (http://forums.trossenrobotics.com/sh...off-some-stuff). Using laser cut acrylic and plastic standoffs for the HROS-1 torso never sat well with me, so intended to replace them with 3D printer torso box to hold all the electronics in a stronger, stiffer, safer design. Unfortunately, I backed a 3D printer on kickstarter instead of buying a kit...

Overall, humanoids are still a bit too specialized compared to wheeled rovers with arms, or even just non-humanoid bipeds. All that extra weight swinging around high off the ground makes control schemes an area of active research while not providing much benefit to non-researchers. A humanoid torso on a wheeled base is much easier for interactions as you don't have to deal with weight, balance, and power/runtime issues anywhere near as severe as with a legged humanoid. A arm/torso-free biped is often adequate for research into bipedal gaits with cost savings on the arms that can go into better servos for the legs.

One of my biggest regrets is that I received a free HROS-1 from Tyberius but never actually did much with it. If only it had gone to a researcher/educator actually capable of being productive instead of me. Ripley has been in a partly assembled state for nearly three years without so much as a minute of actual software development towards making her walk again. I've been trying to make myself partly reassemble her to make a MechWarfare biped (legs with swivel/tilt torso holding airsoft turret), but there are lots of things I've been trying to make myself do for years while failing miserably.

Thank you so much for your detailed response. I was planning to use the kit to get started on the SW side quickly. I will have to check each components to see if I can find all the HR-OS1 frame parts (I have access to the 3D printer, so I can print my own 3D parts), if not I will have to get the Bioloid Premium instead as you have mentioned. Again, thank you for the information.

Kevin

[tician]

What sort of software are you planning to use?

The Bioloid Premium kit is built around Robotis' CM-530 controller which is not compatible with the DARwIn-OP/HROS-1 framework and gait engine. It would allow you to immediately get up and running with a humanoid, but would only be able to use RoboPlus 'Motion' files and 'Task' control programs. Anything more complex than those manually created motions would require reprogramming the CM-530 using custom firmware or replacing the CM-530 with a USB-to-dynamixel converter of some sort connected to a small embedded PC like the RPi. Neither option would be immediately compatible with the DARwIn-OP/HROS-1 frameworks as they are built around the CM-730/Arbotix-Pro which has a few accessories on-board (IMU, buss power control, etc.). The framework basically sends a message to the CM-730 at ~125Hz that gets passed to the servos to update their positions; it also polls the CM-730 for IMU state data and can be made to retrieve servo positions, but cannot remember right now if that was default behavior.

If you are rolling your own software, then it may be least expensive to just buy three or four bulk (6-pack) boxes of AX-12A servos and the metal frames as needed. You then add some custom printed frames for the torso and head pan-tilt mechanism, a 11~12V battery, a USB webcam, a small embedded PC for controlling everything (some version of RaspberryPi or similar), and a USB2AX/U2D2/USB2Dynamixel/Arduino/Teensy/etc. to talk to the dynamixel buss.

I have a spare Arbotix-Pro because I made a cheap shield for the Teensy-3.2 to handle "USB to full-duplex 3.3V UART to half-duplex 5V UART", "USB to full-duplex 3.3V UART to half-duplex 5V RS-485", a 9-DOF IMU/AHRS, and 5V output buffer for Neopixel control. KurtE also rolled his own control board to make something much closer to the CM-730/Arbotix-Pro using a Teensy-3.2 (http://forums.trossenrobotics.com/sh...Teensy-Board-D). Likely others that I am forgetting about right now; my memory sucks.

[tician]

If you are feeling adventurous, it might also be worthwhile to go with the newer XL430 servos (https://www.trossenrobotics.com/dyna...30-w250-t.aspx) (https://www.trossenrobotics.com/dyna...30-w250-t.aspx) instead of AX servos. Slightly more expensive than AX servos since they are not sold in bulk and they use very different frames/brackets, but higher resolution contact-less position sensor, better processor, and software compatible with the other newer X-series servos. The AX servos use a 16MHz ATmega88 with potentiometer and 10-bit ADC for position feedback, while the MX and X servos use 72MHz STM32 with 12-bit magnetic encoder.

...of course, I just remembered you already have the AX-12 servos from a PhantomX, so pointless suggestion.

[drachels]

Hi kshin,
Have you made any progress with this? I have just recently started working again on my HR-SO1 (Joe). Joe is working fine with the original Pi2 but I have been wanting to upgrade to a Pi4. The problems I've run into is the HROS1-Framework is not compatible with the compiler that is now used in the Pi software. Sudo make generates a bunch of errors and lots of warnings. So far I have not been able to get the Arbotix-Pro to respond to anything from the Pi4.

[jwatte]

It may be that the best way to fix that is to learn the language and fix the errors/warnings. They're generally not particularly hard to fix, once you're into it. You could even provide a patch for the open source community!