search

utiasDSL / gym-pybullet-drones

PyBullet Gymnasium environments for single and multi-agent reinforcement learning of quadcopter control
https://utiasDSL.github.io/gym-pybullet-drones/
MIT License
1.18k stars 348 forks source link

About Instantaneously Control of Drones #70

Closed imamim closed 2 years ago

imamim commented 2 years ago

Hi, have a good day

I have a few confusions about instantaneously controlling drones and I want to ask about it. If I need to talk about what I want to reach, it is to be able to control the movements of the drone using the angular and linear velocity values ​​that I set myself. That values are the input, then the needed output is movement.

First of all, when I looked at the past issues, I saw that the same topic was discussed on #49. You mentioned that PyBullet's p.resetBasePositionAndOrientation() and p.resetBaseVelocity() functions can be used here.When I use these functions, do we include the explicit dynamic model that you implement, or do we only use PyBullet dynamics?

Another thing that comes to my mind is; When I examine the examples in the assignments/ folder, a control is processed using obs values obtained from env.step() and then a simulation step is performed with env.step(ACTION). Here ACTION corresponds to engine RPM values. Although the system structure makes sense, it doesn't make much sense for me to go through the RPM calculation.

Should I use BaseAviary.py, ControlAviary.py or control/Control.py instead? How can I control drones as I mentioned above ?

Thanks your answer from now :)

JacopoPan commented 2 years ago

Hello @imamim,

is you question how to instantaneously change/set linear and/or angular velocity to arbitrary values? I would say that PyBullet's p.resetBaseVelocity() method is what you are looking for but, or course, that defies a good part of the purpose of this simulation because you wouldn't adhere anymore to the dynamics of the quadrotor (imposed by its mass, thrust, inertia, torque coefficients, etc.)

imamim commented 2 years ago

Yeah, I understood this issue while working the code yesterday,p.resetBaseVelocity()is working properly. But here was the problem. It was necessary to run the simulationenv.step(action).In the step phase, I always give the action value a value that I set as a constant. Here I observed that this action has no effect after p.resetBaseVelocity, with almost sure (If i'm not escape anything). Also, after giving a random value (and do the step), I get the obs value with the CtrlAviary.py/_computeObs() function.

I am almost certain that these values ​​are correct. The strangeness here was confusing me, but the problem is gone for now, thanks for your answer :)