提出了一种易于脉动阵列实现的平方根椭球状态定界算法。算法将椭球形状矩阵的平方根进行递推计算,使得计算的数值稳定性得以提高。由于平方根算法具有矩阵与矩阵以及矩阵与向量的运算形式,因而适合在并行处理器上执行。为了并行计算,给出了实现此平方根算法的脉动阵列结构。计算复杂性分析显示,若系统状态维数为n,串行计算的计算复杂度至少为O(n^2)。而并行计算的计算复杂度降为O(n)。仿真结果验证了本方法的有效性。
A square-root ellipsoidal state bounding algorithm for parallel computanon using systolic arrays is proposed. For high numerically stability, the Cholesky decomposition is used in the propagation of the shape-defining matrix of the ellipsoid. Since the square-root algorithm is in the form of matrix-matrix and matrix-vector operations, it is suitable for parallel computers. For parallel computation, systolic arrays for the algorithm are developed. Computational complexity analysis reveals that if the system has a state dimension of n, the serial computation is of at least O(n^3 ) complexity per data sample, while the parallel computation is of O(n) complexity. Simulation results show the effectiveness of the proposed approaches.