Neural network for control

I am trying to use brain.jseither synapse.js(or any other JN NN library) to do the following:

  • I have a 3 axis accelerometer
  • I have 4 engines

My input looks like this:

{
  accelX: 0.12,
  accelY: 0.25,
  accelZ: 0.91,
  powerA: 1,
  powerB: 0.75,
  powerC: 0.44,
  powerD: 0.92
}

I want the output to be engine power:

{
  powerA: 1,
  powerB: 0.75,
  powerC: 0.44,
  powerD: 0.92
}

so they are accelapproaching 0. I assume that I can train him by arbitrarily adjusting the power of each engine, and then checking if the acceleration is close to 0. If they are, include him in the training set. If not, do not include in the training kit.

, 7 0 . : ? ANN. , - , , , .

- , , , .

+4
1

, , , . , , , . , - " ", ( ) , . , 0- .

. , , . 100 , . .

. . () . .

. , , , , . Nelder Mead, Genetic Algorithms Hill Climbing.

+1
source

All Articles