Adapt latest crazyflow

[Yuming-Lee24]

I pinned the crazyflow to the latest commit in the pyproject.toml and reproduced the import error on my linux desktop.

commit 1:
now the param loader will be imported from the crazyflow.dynamics and takes in both dynamics model and the drone model. I also changed the signature of build_action_space() to load dynamics.

commit 2:
the params are no longer expand according the world number since learnsyslab/crazyflow#105, a few lines are added in race_core.py to enable it, solved the incompativle shapes for broadcasting error in tests

commit 3:
adapt to the 16 dim of state action. Notice that in race_core.py:

        def apply_action(action: Array, data: EnvData) -> None:
            """Apply the commanded state action to the simulation."""
            ...
            if "action" in disturbances:
                ...
                action += disturbances["action"](subkey, action.shape)
                ...
            return data.replace(sim_data=ctrl_fn(data.sim_data, action))

the disturbance will be added slightly different as before, since it adds on xyzw now.

testing result: all pass on linux and macos, passed the deployment test

[Yuming-Lee24]

the deploy.py did not work, let me check

[Yuming-Lee24]

everything is working now

[ratheron]

Thanks for updating to the latest Crazyflow version, which has some breaking changes.

Added some comments. We should also check the docs for the quat, since only the yaw from the quat is used in the control stack.

[ratheron]

Just one minor thing left for me to approve. Btw, don't we need to change any docs? Seems odd.

[ratheron]

In my opinion, the parameter thrust_max is always per motor, so we shouldn't overwrite it in the dict with 4x. Instead, just take it x4 where we write it into the thrust constraints, as it was before.

[Yuming-Lee24]

yes make sense, but there is somewhere else in the code has ambiguity on total thrust_max and on thrust_max per motor, for example in the build_action_space(), I will suggest to change the variouble name to self.total_thrust_max there.

[Yuming-Lee24]

Just one minor thing left for me to approve. Btw, don't we need to change any docs? Seems odd.

you mean the pr for docs? it haven't done yet, I will finish it tomorrow. Or what do you mean?

[ratheron]

Just one minor thing left for me to approve. Btw, don't we need to change any docs? Seems odd.

you mean the pr for docs? it haven't done yet, I will finish it tomorrow. Or what do you mean?

I mean we change the crazyflow version which broke things. If we mention load_params somewhere in the docs, this also needs to be changed. If the docs never talk about this, it's fine

[Yuming-Lee24]

Just one minor thing left for me to approve. Btw, don't we need to change any docs? Seems odd.

you mean the pr for docs? it haven't done yet, I will finish it tomorrow. Or what do you mean?

I mean we change the crazyflow version which broke things. If we mention load_params somewhere in the docs, this also needs to be changed. If the docs never talk about this, it's fine

I will check the docs, and should I change the variable names to fix the ambiguity in

lsy_drone_racing/lsy_drone_racing/control/attitude_input.py

Lines 41 to 42 in b1f5b36

	self.thrust_min = drone_params["thrust_min"] * 4
	self.thrust_max = drone_params["thrust_max"] * 4

lsy_drone_racing/lsy_drone_racing/envs/race_core.py

Line 239 in b1f5b36

thrust_min, thrust_max = params["thrust_min"] * 4, params["thrust_max"] * 4

[ratheron]

Just one minor thing left for me to approve. Btw, don't we need to change any docs? Seems odd.

you mean the pr for docs? it haven't done yet, I will finish it tomorrow. Or what do you mean?

I mean we change the crazyflow version which broke things. If we mention load_params somewhere in the docs, this also needs to be changed. If the docs never talk about this, it's fine

I will check the docs, and should I change the variable names to fix the ambiguity in

lsy_drone_racing/lsy_drone_racing/control/attitude_input.py

Lines 41 to 42 in b1f5b36

	self.thrust_min = drone_params["thrust_min"] * 4
	self.thrust_max = drone_params["thrust_max"] * 4

lsy_drone_racing/lsy_drone_racing/envs/race_core.py

Line 239 in b1f5b36

thrust_min, thrust_max = params["thrust_min"] * 4, params["thrust_max"] * 4

Yeah, we can remove the ambiguity by saying sth like total_thrust_max