Dynamic model predictive control allocation using CVXGEN

Control allocation deals with the allocation of control among a redundant set of effectors, while taking into account the individual constraints. The use of model predictive control (MPC) for control allocation allows the response times of the actuators to be accounted for, and to take advantage of predictions of the virtual control input as well as differences in dynamic control authority and cost of use among the actuators. Quadratic programming (QP) is essential for implementation of the optimal constrained control allocation strategies. The main contributions of the present paper are the investigation of using the software system CVXGEN and the MPC-based control allocation method. CVXGEN synthesizes a customized portable and library-free C-source code QP interior-point solver for the specific QP problem resulting from the MPC formulation, exploiting structural properties of the specific QP and optimizing the source code for execution speed. Two case studies, one being a missile auto-pilot, illustrates the benefits of using the MPC formulation, and the efficiency of CVXGEN.