Attitude setpoint control

[llanesc]

Currently the only ROS2 interface for the attitude and thrust setpoint is through the cmd_vel_legacy function. However, this option isn't intuitive without looking at the code to figure out the twist to attitude setpoint mapping. Would there be interest in an attitude setpoint ROS2 interface? This would be similar to how I did it in my Crazyswarm2 fork with a custom msg and modifications to the crazyflie_server.py.

[whoenig]

I agree that the naming of cmd_vel_legacy is not good. The issue with that functionality is that parameters can entirely change the behavior, i.e. the values in https://github.com/IMRCLab/crazyflie-firmware/blob/a25e8fb2254f6d5be54ad57566b3b846a97de93b/src/modules/src/crtp_commander_rpyt.c#L226-L275 decide if these commands are actually attitudes or something else.

To me a proper solution would require a firmware change, by adding a additional "generic" type to https://github.com/bitcraze/crazyflie-firmware/blob/master/src/modules/src/crtp_commander_generic.c. Then the behavior is clearly defined and it would make sense to add a custom msg and topic, like we did for some of the other "generic" commands.

What's your take on that @knmcguire ?

[knmcguire]

Yes I would agree that these setpoint behaviors are very legacy....

We could make a PR for the firmware to use 2 ports for the an attitude setpoint and a rate attitude and keep it clean (no parameter magic). Then on the cflib side we can make a commander.send_setpoint_attitude and commander.send_setpoint_attitude_rate.

The only thing is that we have to realize, that the behavior of the current commander.send_setpoint in default is roll, pitch is attitude absolute but yaw is in rate mode (such that the yaw is maintained when the stick is released). That complicates things even further 😄

[llanesc]

In PX4 they seem to provide both absolute and yaw rate setpoint for the attitude setpoint commands. They compute the body rates from the vehicle attitude error and feed forward the yaw rate by transforming to body frame. The reset integral command might also be a good option too because if we switch navigation logic we should also reset the integral (e.g. switching from navigating to hold).

[knmcguire]

Ah oke that is good to know, however we perhaps want to be careful of not trying to accommodate all possibilities but just the ones that people use the most.

So how about for the cflib:

• send_setpoint_manual = thrust percentage, roll/pitch abs attitude, yaw attitude rate
• send_setpoint_manual_rate = thrust percentage, roll/pitch/yaw rate attitude

This implies for what the setpoint is used for instead of what it contains, so it is made for manual flight.

I can make an issue in the cflib for this but I'll wait for after the release, which will happen in these 2 weeks

[llanesc]

Sounds good to me.

[knmcguire]

I've created an issue here: bitcraze/crazyflie-lib-python#461