As you can see, this is just the simulation and if you read the code there are only simulation files, for know im not uploading robot files (which are joystick reading, serial communication for arduino and angle to pulse signal), as i'm working on making them readable and easier to modify for other quadruped robots.
Hi asd, does this program have a fast forward button, for example with fast forward enabled for 5 minutes, the robot would been learning for 50 minutes, or 30 days would be 30 years.
How don't you have thousands of subscribers? Thanks for making it open-source! We could all help to create a cheap and similar alternative to Spot-mini! This is awesome!
Hi Markfps. Not about the Sub. That is true passion, avendture, the big hobby :D nothing can't stop him :D I hope he done with it. Because I want this bot become my pet can run around my home :D we just waiting ~~
@@CinoGear hahahahaha yeah Orion, there is not any intention of earn money or fame, just an unstoppable curiosity on this robot and its mathematical problems involved!
Great project. How would you compare PyBullet and Gazebo? As I have seen from the project files that you have used PyBullet for simulation. I have used Gazebo for quite a time now and I think that Gazebo's simulation environment is hard to set up and handle. Could you give me a summary of the simulation environments used for robotics like Gazebo, PyBullet, MuJoCo, Webots, etc.? Also what works best/widely used in the industry? Thank you :)
Hi, thanks for the good words. I don't have proper experience with other physics engines for these task, some time ago i read a paper that talks just about that (now i can't remenber the name, but you can find it easily on google) it compares gazebo, pybullet, mujoco and some others performance and precision, the conclusions were all of them are really close to reality when well tunned. Thanks for the idea for a future video!
For now i'm just interested in making it stable, as it isn't very robust at walking. Maybe if i can continue with the project in the future i will implement cameras and evironment feedback
@@asdlab Very cool. Have you considered using RL to "learn" the parameters for robust walking? I'm interested because I am just getting started in building a quadruped for a specific task and need to teach it at least object avoidance, navigation, dealing with rough terrain, and a few other semi-autonomous things. Is there a possibility for collaboration and sharing info? I can contribute more on the ML side of things.
What you mean with your AI simulation? You may tell me more about your problem. Are you usong my code? If you are, is it an error? have you tried just to move the debugging slides?
As you can see, this is just the simulation and if you read the code there are only simulation files, for know im not uploading robot files (which are joystick reading, serial communication for arduino and angle to pulse signal), as i'm working on making them readable and easier to modify for other quadruped robots.
Hi asd, does this program have a fast forward button, for example with fast forward enabled for 5 minutes, the robot would been learning for 50 minutes, or 30 days would be 30 years.
How don't you have thousands of subscribers? Thanks for making it open-source! We could all help to create a cheap and similar alternative to Spot-mini! This is awesome!
Thanks a lot man! Maybe it has just started to achive more people and of course open-source is the important thing as we all can contribute.
Hi Markfps. Not about the Sub. That is true passion, avendture, the big hobby :D nothing can't stop him :D I hope he done with it. Because I want this bot become my pet can run around my home :D we just waiting ~~
@@CinoGear hahahahaha yeah Orion, there is not any intention of earn money or fame, just an unstoppable curiosity on this robot and its mathematical problems involved!
Awesome work! Thank you for the detailed descriptions and the code!
Bro bro, that is so amazing, congrat !!!! So perfect project in covid19 boring time
Hahaha yeah, this quarantine helps in the development of the robot
Very nice!
Thanks man!!
I like the gallop gait
Yo, where is the real robot that was in the thumbnail?
Check all the videos before!!
Great project. How would you compare PyBullet and Gazebo? As I have seen from the project files that you have used PyBullet for simulation. I have used Gazebo for quite a time now and I think that Gazebo's simulation environment is hard to set up and handle. Could you give me a summary of the simulation environments used for robotics like Gazebo, PyBullet, MuJoCo, Webots, etc.? Also what works best/widely used in the industry? Thank you :)
Hi, thanks for the good words.
I don't have proper experience with other physics engines for these task, some time ago i read a paper that talks just about that (now i can't remenber the name, but you can find it easily on google) it compares gazebo, pybullet, mujoco and some others performance and precision, the conclusions were all of them are really close to reality when well tunned.
Thanks for the idea for a future video!
@@asdlab Thank you for the reply. You're awesome. Keep making new robots; don't stop. All the very best for your future projects as well :)
thanks for share
Great work. Do you have plans on using RL to do object avoidance, navigation, make autonomous?
For now i'm just interested in making it stable, as it isn't very robust at walking. Maybe if i can continue with the project in the future i will implement cameras and evironment feedback
@@asdlab Very cool. Have you considered using RL to "learn" the parameters for robust walking?
I'm interested because I am just getting started in building a quadruped for a specific task and need to teach it at least object avoidance, navigation, dealing with rough terrain, and a few other semi-autonomous things. Is there a possibility for collaboration and sharing info? I can contribute more on the ML side of things.
nice!
great video, may I ask why my AI simulation only jumping, how I make them walk like dog?
What you mean with your AI simulation? You may tell me more about your problem. Are you usong my code? If you are, is it an error? have you tried just to move the debugging slides?
I have examined files but i have something i cant understand. What is purpose of "kinematic_model.py"? Isn't urdf file enough?
No, .urdf is just for showing the simulation. The kinematic_model.py is the math model that control the movement of every joint.
Hola ese simulador está en los archivos para descargar?!
Si, en github los tienes, si lees el readme sabrás donde esta. Branch V1.
@@asdlab muchas gracias! Exelente trabajo! ahí lo descarge