How can I incorporate Different contrivance structure into linear mixed effect model which is fitrgp /fitlmematrix method I am trying to use?

lme = fitlmematrix(X,y,Z,G,'CovariancePattern','K','FitMethod','REML')

Here the covariance K is for random effect as much as I know. It's a squared covariance kernel. if I take

K=(sigmaF^2)*exp(-0.5*pdist2(x,x','squaredeuclidean')/(sigmaL^2))

it exceeds the dimension of covariance pattern in fitlmematrix. Here my error variance would be ~N(0,sigma^2I). So how can I incorporate Gaussian correlation function here?