reinforcement learning, 3D simulink model

how to built a reset function of your 3D simulink model in reinforcement learning? i am getting state from rectangular joint in model now after every episode i have to reset the simulink model if my requirement have not been meet those requirment is based on state. my question is how can i reset the simulink model in model script?

Answer 1

[Emmanouil Tzorakoleftherakis]

There is nothing specific you need to do for a 3d Simulink model. You can follow any other Simulink example from Reinforcement Learning Toolbox. One example that sounds relevant is this one. I would take a look at the 'walkerResetFcn' script.

[Fawad Salam]

thanks for replying. walkerResetFcn is based on initial condition of simulink model of walking robot but in my case i am working on quadrotor and the value i am getting is from joints used in simulink model of qaud rotor. now as you now during trainning after every episode you need to reset the model in order to move the quadrotor in the direction. do you know how can i reset the model?

[Emmanouil Tzorakoleftherakis]

I am not really seeing how this is any different. The reset function I mentioned above randomizes the angles and these values are then used as initial conditions in the respective joint blocks in the robot model.

[Fawad Salam]

why he is randomizing the angle ? can you not simply just reset the joints on the bases of output state value?

like

if(X=0 && y=0 )

reset;

else

notreset

end

reset mean you initialize the vx0 vy0 to zero

?

[Emmanouil Tzorakoleftherakis]

If you don't reset the joint angles, you may be overfitting to the initial condition of the episode. At the end of the day this is optional of course, it's up to you to specify what you want to reset and how.

[Fawad Salam]

how can i used the ouput state signal coming from joint in if else comparator check? because i want to reset the model basics on the output state value.