What I want to achieve is to build a neural network with Keras which will have to stabilize a quadcopter I've built. The network would have three inputs: pitch, roll, yaw, acceleration x, accel y and accel z and four outputs: pwm length for motor 1, 2, 3 and 4. The problem is that the length of the pwm wave needs to fit the range from 1000 (no throttle) to 2000 (max throttle). I know I can map the output to this range, but will this work? And another problem is if it's even possible to train such a network, because I have no model data, they would be submitted real-time during training, when the mpu6050 updates the gyro and accel data, which will rely on the network's output - the speed of each motor?