N = ccl_learna_pred_proj_alpha (model, q, Iu) Prediction of the projection matrix Our model predicts the constraint parameters. this function is used to reconstuct the projection matrix from constraint paramters. Input: model Parametric model parameters q Joint state data Iu Identity matrix Output: N Null space projection
0001 function N = ccl_learna_pred_proj_alpha (model, q, Iu) 0002 % N = ccl_learna_pred_proj_alpha (model, q, Iu) 0003 % Prediction of the projection matrix 0004 % Our model predicts the constraint parameters. this function is used to 0005 % reconstuct the projection matrix from constraint paramters. 0006 % 0007 % Input: 0008 % model Parametric model parameters 0009 % q Joint state data 0010 % Iu Identity matrix 0011 % 0012 % Output: 0013 % N Null space projection 0014 0015 Rn = Iu ; % initial rotation 0016 A = zeros(model.dim_k, model.dim_u) ; % Initial constraint matrix 0017 bx = model.phi(q) ; % gaussian kernel of q 0018 for k = 1:model.dim_k 0019 theta = [ pi/2 * ones(1,k-1), (model.w{k} * bx )' ] ; % the kth constraint parameter 0020 A(k,:) = ccl_math_uvec(theta) * Rn ; % the kth constraint vector 0021 Rn = ccl_math_rotmat (theta, Rn, model, k) ; % update rotation matrix for (k+1) 0022 end 0023 N = Iu - A'*A ; 0024 end