Its not just a stabilization/rate control loop but the NN does both guidance and control. It is an MLP that takes state + upcoming gate positions/waypoints and controls the motors directly. I.e. there is no explicit…
Not much since it doesn't take images as it's input and it's running on an embedded mcu. Based on the papers linked elsewhere in here the state vector is 24 floats and it's running at 1kHz so around 100 kB/s.
No, but there's a previous paper for the controller used: https://arxiv.org/abs/2504.21586
It's fundamentally different, it's using an RL trained network that gets the drone state (position, orientation, velocity) as input and directly outputs motor commands.
It used a small RL trained network running on the flight controllers MCU directly that controlled the motors given state (position, orientation ...) inputs. The Jetson handled vision processing.
Its not just a stabilization/rate control loop but the NN does both guidance and control. It is an MLP that takes state + upcoming gate positions/waypoints and controls the motors directly. I.e. there is no explicit…
Not much since it doesn't take images as it's input and it's running on an embedded mcu. Based on the papers linked elsewhere in here the state vector is 24 floats and it's running at 1kHz so around 100 kB/s.
No, but there's a previous paper for the controller used: https://arxiv.org/abs/2504.21586
It's fundamentally different, it's using an RL trained network that gets the drone state (position, orientation, velocity) as input and directly outputs motor commands.
It used a small RL trained network running on the flight controllers MCU directly that controlled the motors given state (position, orientation ...) inputs. The Jetson handled vision processing.