While scanning the Poppy website daily for news of the current priority of poppy development, the current priority remains unclear. Are you currently working on autonomous walking capability?; or something else? Once walking is achieved, many more roboticists will join your effort!
Our priorities are more focused around building a robust and easy to use robotic platform. In this context, @Fdepraetre is making a whole new electronic architecture arduino-based and we are correcting some non-optimized mechanic parts and assembly processes.
Another priority is the community management. We are looking toward efficient solution to organise contributions and simplify the access to Poppy i.e. getting hardware stuff and understanding how to make it work. Currently, this aspect is not well designed and it would maybe require the help of contributors to make a better wiki, a better main website (with updated news) and a more clear documentation.
Of course, walking and balancing are a main mid-term goal. @Steve is actually working on the balance while @Alexandre is working on the foot design to simplify both walking and balancing control.
There are also a lot of welcomed extension features we did not find time to develop yet such as:
- Binding ROS for pypot
- Integration of Poppy in a physical simulator such as vrep or gazebo
- Graphic interface to monitor the robot sensors
- A Poppy101 code for new user to test Poppy easily.
- Adaptation of the Poppy parts to low cost servomotors
- Design of expressive and emotive eyes to be displayed on the LCD screen
- Design of accessories such as bag, chair, support, …
- And many more …
Also all these “welcomed contributions” should be listed somewhere to suggest work for potential interested contributors but it has not be done yet …
Are you committed to developing your own walking controllers? Or will this also be an area open to community participation? I’m very interested in open source robotics, I’ve been working on a controller for balance and walking, I wonder if it could be useful in your project?
If interested, here’s a 30 second video of a simulation I made. In it, you can see a biped model stand, balance, push off and take 12 steps before falling. Needs a lot of work, but I think it gets the idea across :
The walking and balancing behavior are definitely an open area for the community. We will certainly develop a walking controller but not The walking controller. In our lab we are interested by some specific ways to achieve sensorimotor skill (most of them are epigenetic) and we do not impose them.
We would be of course interested by any good idea on biped locomotion.
The simulation you present is really interesting, maybe you can share more about your research in the science category ?
Thank you, I’ll post more information there.
I really like your simulation work!
1–What simulator are you using?
2–Can you share your controller code? I’m not familiar with what data a simulator must have to accurately simulate Poppy, but it seems that the dimensions and material density of the Poppy body as well as how much all parts weigh individually must be included?!
3–Does the simulator already have weight allocations for various types of materials?
4–Are you using an evolutionary algorithm to arrive at a working walking gait?
5–Are you using a walking gait generated by “not” using ZMP?
To answer your questions:
- Matlab & Simmechanics
- I’ll be putting the source code and model description online, I’ll put references in a later post within the Science section.
3)Each segment you see has weight and density properties. The model was not designed to match the material properties of Poppy, I didn’t know about it when I started. But it was designed to experiment with different mass distributions. For instance, the mass, density and position of those yellow cubes can be varied. But I think some insights gathered from work on this model could be useful with Poppy - The gait is simply based on human kinesiological data, there was some trial and error involved to account for the lack of feet. I’ll post details later.
5)This model uses ideas from the fields of dynamic walking and intermittent control. It uses the passive dynamics of the system as much as possible and only applies control signals at brief intervals. The idea is to minimize power, signal bandwidth and computing resources, thereby reducing cost.
This is very much a work in progress, click my user profile to find my later posts.
1 Like