Long story short, I am a computer engineering student in his senior year. My senior capstone project requires me to build an autonomous drone with deep machine learning. I stumbled across PX4 and pixhawk in my brief work in the aerospace club and I was thinking about maybe incorporating PX4 into my senior capstone project. However, from what it looks like, PX4 autopilot does not utilize any machine learning but rather PX4 utilizes predetermined set parameters with gps in the QGroundControl. Can someone please help verify this for me? I am rather new and perusing through this system.
That’s correct, and it’s also true of all other flight controllers. However, PX4 and the Pixhawk hardware are designed to work with a “companion computer (CC)” (typically a Linux-based board like a Raspberry Pi) which would handle higher-level functions such as camera image processing, which is where deep learning is typically done. So you would use use deep learning on the CC to identify and track an object, for example, and then send commands to PX4 to follow it
Thank you so much! That clarifies a lot. I was planning on retrofitting a drone with raspberry pi and implement my machine learning onto the microcontroller as well.
