Reading Forces & Accelerometer from PyBullet

[hammoudbilal]

I am writing this as a reference as I might forget about it by next week. I will fix it in the next few days, but here are some issues,

1. PinBulletWrapper.get_force() seems to return contact forces in the negative direction (clear on the z axis not so sure about the other directions). Would it make sense to flip it? I will look into pybullet documentation and why is this happening.

2. The other issue with PinBulletWrapper.get_force() is that it returns only the active contacts, so we get a varying size list, would it make sense to store this list in the fixed number of end effectors, and set all in-active contacts to zero? or will that break too much code ? @jviereck

3. @nrotella as I understand you added all the IMU functionality to the same wrapper? Just to confirm PinBulletWrapper.get_base_imu_linacc() and PinBulletWrapper.get_base_imu_angvel() transform the IMU readings from IMU frame to base frame or is it the other way arround? but in order to do that I have to call PinBulletWrapper.compute_numerical_quantities(dt) first ?

let me know what you guys think and I'll work on fixing/changing stuff if need be :)

[jviereck]

1. Sure
2. Happy to change this as well.
3. How about we call PinBulletWrapper.compute_numerical_quantities(dt) with the step function?

[hammoudbilal]

I can take care of the first two points, and it seems that PinBulletWrapper.compute_numerical_quantities(dt) gets computed in BulletEnv.step() so no need to take action on this one. I will close this issue once I fix the first two points