A fast algorithm for MPC based on aggregation strategy

Improving the computational efficiency for model predictive control (MPC) becomes the focus of recent researches. This paper proposes a fast algorithm to solve the quadratic programming (QP) problem of MPC for systems with only input constraints. This algorithm improves the efficiency of the fast gradient approach. Inspired by the multiplexed way, this algorithm searches a better solution of QP problem along the direction of one input channel in every iteration. Meanwhile, in order to further reduce the online computational complexity, the proposed algorithm applies the aggregation strategy to each input channel. Due to the aggregation strategy and the determined optimization direction, the proposed algorithm reduces the number of vector multiplications of each iteration, i.e. with less online computational complexity than previous works. Applying the proposed algorithm, simulations for a system with 3 inputs show that computational efficiency of this algorithm can reach a level of tens of microseconds in Matlab environment.