damp_gaussianhills.m

approximate reinforcement learning

function b = damp_gaussianhills(gh, x)
% Damping variation as a function of the position.
%   B = DAMP_GAUSSIANHILLS(GH, X)
% `Gaussian hills': the damping is a sum of gaussians in the 2-dimensional area with given
% centers, spreads, and amplitudes.
%
% Parameters:
%   GH  - the parameters of the gaussian hills, a structure with fields
%       n       - the number of gaussians in the friction landscape
%       a       - the amplitude (height) of the gaussians, an array with n elements
%       c       - the centers of the gaussians, 2xn matrix, each column a center point
%       sigma   - the spread of the gaussians, 2xn matrix, each column coordinate-wise spread
%   X   - the current state, position should be on the first two elements
% Returns:
%   B   - the friction (damping) coefficient in X


pos = repmat(x([1;2]), 1, gh.n);
b = gh.b0 + sum( gh.a .* exp(-sum( ((pos - gh.c) ./ gh.sigma) .^ 2 )) ); 


% END damp_gaussianhills, RETURNING b ===================================================