Institut de Robòtica i Informàtica Industrial

DiscretizeRewardModel

PURPOSE ^

Discretizes the reward model on the action side.

SYNOPSIS ^

function DRM=DiscretizeRewardModel(RM,A)

DESCRIPTION ^

   Discretizes the reward model on the action side.

   Defines a reward function on continuous states and discrete actions 
   from a reward function on continuous states and actions.
   The provided set of 'samples' in the action space is used as the set of
   actions in the discrete set to be defined.

   Parameters:
     RM: The continous state/action action model to be discretized.
     A: The discrete set of actions to use in the discretization.

   See also CS_DA_RewardModel.

CROSS-REFERENCE INFORMATION ^

This function calls:
  • GMixture Gaussian mixture constructor.
  • Value Evaluates a GMixture.
  • Value Evaluation of a Gaussian.
  • CS_DA_RewardModel CS_DA_RewardModel constructor.
  • dim Dimensionality of a continuous space.
  • dim Dimensionality of a discrete space.
This function is called by:

SOURCE CODE ^

0001 function DRM=DiscretizeRewardModel(RM,A)
0002 %   Discretizes the reward model on the action side.
0003 %
0004 %   Defines a reward function on continuous states and discrete actions
0005 %   from a reward function on continuous states and actions.
0006 %   The provided set of 'samples' in the action space is used as the set of
0007 %   actions in the discrete set to be defined.
0008 %
0009 %   Parameters:
0010 %     RM: The continous state/action action model to be discretized.
0011 %     A: The discrete set of actions to use in the discretization.
0012 %
0013 %   See also CS_DA_RewardModel.
0014 
0015   na=dim(A);
0016   
0017   r=cell(1,na);
0018   for i=1:na
0019     r{i}=GMixture(RM.w.*cellfun(@(ga)(Value(ga,A(i))),RM.gA),RM.gS);
0020   end
0021   
0022   DRM=CS_DA_RewardModel(RM.S,A,r);


Institut de Robòtica i Informàtica Industrial

Generated on Wed 05-Aug-2009 15:05:21 by m2html © 2003