Accelerating sequential least squares active set control allocation

This is the authors' abstract. We don't add key points for this paper.

The Sequential Least Squares Active Set (SLS-AS) algorithm is commonly used in control allocation (CA) of safety critical systems. The SLS-AS is computationally more expensive compared to the Weighted Least Squares (WLS), which is its major drawback. This paper proposes modifications to the SLS-AS, reducing the computational effort by 50 % compared to the original implementation while retaining the effectiveness. The accelerated version is compared to the Redistributed Scaled Pseudoinverse (RSPI) with Null-Space Transition (NST), which is related to SLS-AS, using randomly generated data based on real systems as well as real system data. Those experiments show that both algorithms perform similarly well. Further, the paper evaluates the effect of aborting the algorithm before it converges, to reduce the execution time. The work demonstrates that both algorithms are suitable for real-time CA.