search

utiasDSL / gym-pybullet-drones

PyBullet Gymnasium environments for single and multi-agent reinforcement learning of quadcopter control
https://utiasDSL.github.io/gym-pybullet-drones/
MIT License
1.16k stars 340 forks source link

[question] difference of crazyflie's kf parameter in scientific literature? (urdf file) #146

Open hocaso opened 1 year ago

hocaso commented 1 year ago

I am checking the parameters of the crazyflie used in https://github.com/utiasDSL/gym-pybullet-drones/blob/master/gym_pybullet_drones/assets/cf2x.urdf

The references of the file point to 3 urls, the middle one doesn't work (Julian) right now, but checking the first and third (Carlos and MIT) I realized that the kf parameters are quite different even when converting to the same units. If I'm not mistaken, the kf values are 3.16e-10 N*rpm**2 (the one used by the library, denoted as Ct, page 12 of the document) and 0.005022 N*s**2 (from MIT, page 39). When converting this last one to compare with the first one I obtain 5.5e-5 N*rpm**2.

It seems the magnitude is square from one another, is there something I'm missing? From the formulae used in the documents they seem to denote the same thing...

JacopoPan commented 1 year ago

Hi @hocaso

thanks for highlighting the broken link, I fixed it here https://github.com/utiasDSL/gym-pybullet-drones/blob/2f1e136351255e1b920cce4a8637e45d4fff445b/gym_pybullet_drones/assets/cf2x.urdf#L17 and yes, the basic modeling used here is based on Carlos' work (who was a DSL member) and performed well transferring to the real system. I can look a bit deeper into this if you think there is a mistake. You can also double check with the results on Bitcraze's website

video