Hi,
I am thinking of implementing neural network on crazyflie 2.0. Could any one of you help me in guiding how to approach to achieve this?
Thanks in advance.
Neural network for crazyflie 2.0
Re: Neural network for crazyflie 2.0
Hi,
The answer depends on what you are trying to achieve and what are your interests.
If you want to control fast things like the Crazyflie orientation I guess the only way it to program your neural network in the firmware itself.
If you have access to some sort of local positioning system and are more interested in high level control like position/behavior it should be enough to add the network on the ground and control high level behavior from there, at least for the beginning.
If you have a more precise idea of what you want to control with a neural network I can be a bit more precise.
/Arnaud
The answer depends on what you are trying to achieve and what are your interests.
If you want to control fast things like the Crazyflie orientation I guess the only way it to program your neural network in the firmware itself.
If you have access to some sort of local positioning system and are more interested in high level control like position/behavior it should be enough to add the network on the ground and control high level behavior from there, at least for the beginning.
If you have a more precise idea of what you want to control with a neural network I can be a bit more precise.
/Arnaud
Re: Neural network for crazyflie 2.0
I've managed to calculate the motorratios on my pc and send them then to the crazyflie (1.0 version). But I couldn't use your logging routine. Instead every time a packet with new motorratios arrive at the crazyflie I immediately send a packet containing the euler- and gyro-data via crtpSendPacket back to my pc.arnaud wrote:If you want to control fast things like the Crazyflie orientation I guess the only way it to program your neural network in the firmware itself.
Re: Neural network for crazyflie 2.0
Interesting, does it still fly with external attitude control?
Re: Neural network for crazyflie 2.0
Yes it does. I also realised a position control with two LQ-regulators. The outer regulator calculates the desired roll, pitch and thrust. The inner one the necessary motorratios. Using a stereo camera I determine the position of the crazyflie. As soon as I upload a video I can post a link to the video.
Re: Neural network for crazyflie 2.0
Sounds great, I can't wait for the video .
Re: Neural network for crazyflie 2.0
Videolink is in a new thread: viewtopic.php?f=11&t=2094