XPost: comp.ai.neural-nets   
   From: thorstenkiefer@gmx.de   
      
   Hi,   
   I modeled a Robot using ODE(Open Dynamics Engine). Now I want it to stand   
   up. But I have no algorithm for that. So I want it to learn how to stand   
   up. The neural net has as inputs the absolute positions of all limbs, and   
   additionally the desired action to take (i.e. the force to be applied to   
   the hinges), and it should compute the value of that state-action pair. A   
   low vlaue means "good" and a high value means "bad". So I try out several   
   force-sets (while the state keeps constant) and take that force-set for   
   which the NN computed the lowest value. By the time the NN should learn how   
   to compute the correct value for a given state-action-pair. This is   
   something like Reinforcement learning, but it doesn't work. Does anyone   
   have an idea how to make the robot learn how to stand up ?   
      
   Regards   
   Thorsten   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)