The first problem for direct thrust control is thrust estimation.A wrapper algorithm

which reduces the computational complexity and also enhances the model generalization performance

was proposed based on least squares support vector regression machine.In addition

as a surrogate of traditional covariance method

QR decomposition was utilized to build sparse model strengthening the numerical stability.Finally

a thrust estimator design was applied to validate the effectiveness and feasibility of the proposed feature selection algorithm and sparse model using QR decomposition.