CS_DO_ObsModel
PURPOSE 
CS_DO_ObsModel constructor.
SYNOPSIS
function OM=CS_DO_ObsModel(varargin)
DESCRIPTION
CS_DO_ObsModel constructor.
Defines an observation model (p(o|s)) on continuous state and discrete
observation spaces.
Parameters
S: Continuous state space.
O: Discrete observation space.
p: A cell array with one GMixture (in 's') for each observation.CROSS-REFERENCE INFORMATION
This function calls:
- size Returns the size of a policy.
- rand Random state from a discrete belief.
- rand Random state from a belief.
- Value Evaluates a GMixture.
- rand Generates random points on a GMixture.
- Value Evaluation of a Gaussian.
- rand Generates random ponts on a Gaussian.
- ObsModel ObsModel constructor.
- UniformProbability Uniform probability value on a continuous space.
- dim Dimensionality of a continuous space.
- rand Random state from a continuous space.
- UniformProbability Uniform probability value on a discrete space.
- dim Dimensionality of a discrete space.
- rand Random state from a discrete space.
- GetTest1Parameters The example on Figure 1.
- DiscretizeObsModel Discretizes the observation model on the observation side.
SOURCE CODE
0001 function OM=CS_DO_ObsModel(varargin) 0002 % CS_DO_ObsModel constructor. 0003 % 0004 % Defines an observation model (p(o|s)) on continuous state and discrete 0005 % observation spaces. 0006 % 0007 % Parameters 0008 % S: Continuous state space. 0009 % O: Discrete observation space. 0010 % p: A cell array with one GMixture (in 's') for each observation. 0011 0012 switch nargin 0013 case 1 0014 if isa(varargin{1},'CS_DO_ObsModel') 0015 OM=varargin{1}; 0016 else 0017 error('Wrong parameter type in CS_DO_ObsModel constructor'); 0018 end 0019 case 3 0020 if isa(varargin{1},'CSpace') 0021 OM.S=varargin{1}; 0022 else 0023 error('Wrong parameter type in CS_DO_ObsModel constructor'); 0024 end 0025 if isa(varargin{2},'DSpace') 0026 OM.O=varargin{2}; 0027 else 0028 error('Wrong parameter type in CS_DO_ObsModel constructor'); 0029 end 0030 if isa(varargin{3},'cell') 0031 OM.p=varargin{3}; 0032 else 0033 error('Wrong parameter type in CS_DO_ObsModel constructor'); 0034 end 0035 0036 no=dim(OM.O); 0037 if no~=size(OM.p,2) 0038 error('Size missmatch in DS_CO_ObsModel constructor'); 0039 end 0040 0041 % p(s) should be uniform in 's'. We compute p(s) as 0042 % p(s) = sum_o p(o,s) 0043 % and, assuming it is constant, we scale p(o,s) so that p(s) is 0044 % actually uniform. 0045 o=OM.p{1}; 0046 for i=2:no 0047 o=o+OM.p{i}; 0048 end 0049 % Scale factor to apply = What it should be (UniformProbability(OM.S)) 0050 % vs. what actually is (Value(o,Center(OM.S))) 0051 % We estimate p(s) via sampling (small variations can occur) 0052 ps=0; 0053 for i=1:10 0054 ps=ps+Value(o,rand(OM.S)); 0055 end 0056 ps=ps/10; 0057 scale=UniformProbability(OM.S)/ps; 0058 for i=1:no 0059 OM.p{i}=OM.p{i}*scale; 0060 end 0061 0062 OMBase=ObsModel(); 0063 0064 OM=class(OM,'CS_DO_ObsModel',OMBase); 0065 0066 otherwise 0067 error('Wrong number of parameters in CS_DO_ObsModel constructor'); 0068 end 0069 0070