Drone or UAV Platform for deployment of Deep Learning / Reinforcement Learning

Question

Good day,

It has been a while since I last posted. I would just like to ask if anyone part of a lab or company doing research on aerial robotics has any suggestions of a good platform for deploying computer vision algorithms for aerial robots?

Currently our lab has a set of DJI Matrice drones but are too heavy for our liking. We really wanted to use the Skydio R2 drone for out future research projects but found out later that the SDK does not allow access for implementing our own Deep Learning or Reinforcement Learning networks (only allows the user to code their own preset movements in python). We also took a look at the Parrot AR drones but found that they were discontinued and do not have the computing power that the Skydio has.

I suggested that we stick with the DJI Matrice and just use an Nvidia Jetson for deployment but I am still curious to know if anyone knows of other available platforms. Does anyone know of a more compact platform available for research purposes?

Thank you :)

Answer 1

there seems to be no mature drone platform focused on the NN research

personally I suggest using a drone combines a NUC, TX2 for example

1. the drone works as a flying platform providing control APIs, and a gimble system if possible
2. the vision system can be constructed by NUC, you can put your own camera on it and do your NN research

for this design, you need a drone platform that should be friendly to mount customize devices. I suggest Matrices 100 or DJI A3 with your own frame, which have enough space and can privide power to your NUC.