Dog Control

All the stuff you need to control a robot dog.

TODO

Goal : having a controller with low PD gains that keeps the robot upright both in simulation and in real life.

• faire le bout de code qui transforme la force désirée au niveau d'un pied à une target en position. -> done
  • On a : longueur vectoriel force égal moment (cross(L,F)=M) pour tous les joints -> done
  • On projette le moment dans ce que peux faire le joint. On clip le moment dans un truc que le joint peut faire. -> done
  • On déduit un delta de position target (en prenant en compte le kp différent pour chaque joint) -> done
• solve a least-square problem to compute the forces at the feets needed to put the dog back into balance (QP problem, use quadprog) (need to think about what we want as force / torque applied on the main body) -> work in progress
  • calculer ce qu'on veut comme force/torque en fonction de l'orientation/vitesse de rotation du torse
• add the constraints that the forces must be over the ground
• add the constraint that the torques at each motor must be smaller than some threshold (measure the max torque for each motor)

Notes

mujoco

mujoco xml reference : https://mujoco.readthedocs.io/en/latest/XMLreference.html#xml-reference

And the other very good one : https://mujoco.readthedocs.io/en/latest/APIreference.html#mjdata

Jacobian explanation : https://sci-hub.hkvisa.net/10.1016/0094-114x(83)90089-7

raspberry

To setup a network on the raspberry :

• sudo raspi-config
• System option
• Wireless LAN
• Enter Box name and password