I was a nanodegree student in Spring 2018 so wondering what changes were made if any. I know there was some concern about parts of physics_sim being incorrect: https://discussions.udacity.com/t/error-in-physics-sim-py/802561

Please clarify as I am redoing this project's three tasks so I can try to string them together for a take-off -> hover -> land scenario.

I posted a question on Datascience Stackexchange: https://datascience.stackexchange.com/questions/47526/what-are-some-ways-to-render-a-ddpg-autonomous-quadcopter-like-openai-gym

What are some ways to do this?

Is this (or RL-Quadcopter) still taught in any Udacity course?

Hi Alexa

I have placed the quadcopter physics into AI Gym. I shall include a GitHub repo link to that in a subsequent post. I have checked the physics lines of your code with mine and there is no difference. That repo link I'll send has the notebook to test the AI gym quadcopter.

I have noticed a velocity divergence for negative x and y body frame velocities. I'm a bit worried as this may be affecting the RL-Quadcopter project, (although I imagine the RL can compensate if properly trained).

Test:

In level flight at height of 10, zero angles:

I initialize the velocity to vx=10 and that appears to damp away.

I initialize the velocity to vx=-10 and that appears to damp away.

When I was experimenting, I found a bug in PhysicsSim.
It doesn't work with init_velocities properly. please check the following sample.
I will create a pull request for this bug.

# init
sim = PhysicsSim(init_velocities=np.array([0.0, 0.0, 0.0]))
sim.v
=> array([0., 0., 0.])
sim.init_velocities
=> array([0., 0., 0.])

# take step
sim.next_timestep([0.0, 0.0, 0.0, 0.0])

sim.init_velocities
=> array([ 0.    ,  0.    , -0.1962]) # init_velocities has been changed

sim.reset()

sim.v
=> array([ 0.    ,  0.    , -0.1962]) # initial v didn't get reset properly due to the updated init_velocities