因此，智能体的移动序列是2-3-1-5。

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Q learning of single agent move in N rooms
% Matlab Code companion of
% Q Learning by Example
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function q=ReinforcementLearning
clc;
format short
format compact

    % Two input: R and gamma
    % immediate reward matrix;
    % row and column = states; -Inf = no door between room
    R=[-inf,-inf,-inf,-inf,   0, -inf;
       -inf,-inf,-inf,   0,-inf, 100;
       -inf,-inf,-inf,   0,-inf, -inf;
       -inf,   0,   0,-inf,   0, -inf;
          0,-inf,-inf,   0,-inf, 100;
       -inf,   0,-inf,-inf,   0, 100];

    gamma=0.80;            % learning parameter

    q=zeros(size(R));      % initialize Q as zero,q的行数和列数等于矩阵R的。
    q1=ones(size(R))*inf;  % initialize previous Q as big number
    count=0;               % counter
    
    for episode=0:50000
       % random initial state
       y=randperm(size(R,1));%产生1到6的随机数%a=size(R,1)把矩阵R的行数返回给a,b=size(R,2)把矩阵R的列数返回给b
       state=y(1);           %取1到6的随机数的第一个数
       
       % select any action from this state
       x=find(R(state,:)>=0);        % find possible action of this state.返回矩阵R第state行所有列中不小于零的数据的下标
       if size(x,1)>0,
          x1=RandomPermutation(x);   % randomize the possible action
          x1=x1(1);                  % select an action
       end

       qMax=max(q,[],2);
       q(state,x1)= R(state,x1)+gamma*qMax(x1);   % get max of all actions
   
       
       % break if convergence: small deviation on q for 1000 consecutive
       if sum(sum(abs(q1-q)))<0.0001 & sum(sum(q >0))
          if count>1000,
             episode        % report last episode
             break          % for
          else
             count=count+1; % set counter if deviation of q is small
          end
       else
          q1=q;
          count=0; % reset counter when deviation of q from previous q is large
       end
    end

    %normalize q
    g=max(max(q));
    if g>0,
       q=100*q/g;
    end
      


   % The code above is using basic library RandomPermutation below

 
 function y=RandomPermutation(A)
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % return random permutation of matrix A
 % unlike randperm(n) that give permutation of integer 1:n only,
 % RandomPermutation rearrange member of matrix A randomly
 % This function is useful for MonteCarlo Simulation,
 %  Bootstrap sampling, game, etc.
 %
 % Copyright Kardi Teknomo(c) 2005
 % (http://people.revoledu.com/kardi/)
 %
 % example: A = [ 2, 1, 5, 3]
 % RandomPermutation(A) may produce [ 1, 5, 3, 2] or [ 5, 3, 2, 3]
 %
 % example:
 % A=magic(3)
 % RandomPermutation(A)
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
   [r,c]=size(A);
   b=reshape(A,r*c,1);       % convert to column vector
   x=randperm(r*c);          % make integer permutation of similar array as key
   w=[b,x'];                 % combine matrix and key
   d=sortrows(w,2);          % sort according to key
   y=reshape(d(:,1),r,c);    % return back the matrix