Improving the performance of active set based Model Predictive Controls by dataflow methods

Dataflow representations of Digital Signal Processing (DSP) software have been developing since the 1980's. They have proven to be useful in identifying bottlenecks in DSP algorithms, improving the efficiency of the computations, and in designing appropriate hardware for implementing the algorithms. This paper demonstrates the use of dataflow to improve a Model Predictive Control (MPC) algorithm. MPC has been extensively used in the real world but its application has been limited to relatively slow processes because it is computationally intensive. It is shown that the time required for the MPC computations using a representative active set method for solving the optimization problem can be reduced by means of dataflow analysis and implementation improvements based on these analyses.