From: Robust direct position determination methods in the presence of array model errors
Calculation module | Computational unit | Computational complexity of each unit | Total computational complexity |
---|---|---|---|
Gauss-Newton algorithm in (56) | Calculation of \( {t}_{0,m}\left({\widehat{\mathbf{p}}}^{(i)}\right) \) | \( {\displaystyle \begin{array}{c}{MN}^2{K}^2+\left(N+2\right){K}^2\\ {}+0.5 NK\cdot {\log}_2K\end{array}} \) | \( {I}_{\mathrm{iter}}\left(\begin{array}{c}{I}_{\mathrm{N}}\left(\begin{array}{c}{MN}^2{K}^2+{N}^2\left(3{K}^2L+2{K}^2\right)\\ {}+N\left(\begin{array}{c}{KL}^3+{KL}^2+2 KL\\ {}+{K}^2+K+0.5K\cdot {\log}_2K\end{array}\right)\\ {}+6{K}^2+4K+2{L}^2+O\left({L}^3\right)\end{array}\right)\\ {}+{I}_{\mathrm{G}\hbox{-} \mathrm{N}}N\left(\begin{array}{c}{Q}_n^2\left( MK+3\right)\\ {}+{Q}_n\left({M}^2\left({K}^2+K\right)+2 MK\right)\\ {}+{M}^2\left(3{K}^2+K\right)+M\left(2K+1\right)\\ {}+2K+O\left({Q}_n^3\right)\end{array}\right)\\ {}+M\left({N}^2{K}^2+2 NK+N\right)+\left(N+2\right){K}^2\\ {}+0.5 NK\cdot {\log}_2K+ NK\end{array}\right) \) (where IN, IG ‐ N, and Iiter denote the iteration numbers for Gauss-Newton algorithm and alternating minimization algorithm, respectively) |
\( \mathbf{R}\left({\widehat{\mathbf{p}}}^{(i)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right) \) | \( {\displaystyle \begin{array}{c}{N}^2\left(2{K}^2L+2{K}^2\right)\\ {}+N\left(\begin{array}{c}{KL}^3+{KL}^2\\ {}+ KL+K\end{array}\right)\\ {}+3{K}^2+3K+2{L}^2\end{array}} \) | ||
\( \mathbf{r}\left({\widehat{\mathbf{p}}}^{(i)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right) \) | \( {\displaystyle \begin{array}{c}{N}^2{K}^2L+ NKL\\ {}+{K}^2+K\end{array}} \) | ||
\( {\alpha}^i{\left(\mathbf{R}\left({\widehat{\mathbf{p}}}^{(i)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right)\right)}^{-1}\mathbf{r}\left({\widehat{\mathbf{p}}}^{(i)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right) \) | O(L3) | ||
Estimation of t0 | \( \mathbf{B}\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right) \) | MN 2 K 2 | |
\( {{\overline{\mathbf{S}}}^{\prime}}^{\mathrm{H}}\mathbf{B}\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right){\left(\mathbf{B}\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{\boldsymbol{\upmu}}}^{\left(\mathrm{b}\right)}\right)\right)}^{\mathrm{H}}{\overline{\mathbf{S}}}^{\prime } \) | (N + 2)K2 | ||
FFT algorithm | 0.5NK ⋅ log2K | ||
Gauss-Newton algorithm in (33) | \( {\mathbf{g}}_n\left({\widehat{\boldsymbol{\upmu}}}_n^{(i)}\right) \) | \( {\displaystyle \begin{array}{c}{Q}_n^2+2{M}^2{K}^2\\ {}+M\left(K+1\right)+2K\end{array}} \) | |
\( {\mathbf{G}}_n\left({\widehat{\boldsymbol{\upmu}}}_n^{(i)}\right) \) | \( {\displaystyle \begin{array}{c}{Q}_n\left(\begin{array}{c}{M}^2\left({K}^2+K\right)\\ {}+ MK\end{array}\right)\\ {}+{M}^2\left({K}^2+K\right)+ MK\end{array}} \) | ||
\( {\displaystyle \begin{array}{c}{Q}_n^2\left( MK+2\right)\\ {}+{Q}_n MK+O\left({Q}_n^3\right)\end{array}} \) | |||
Estimation of β | \( \frac{{\left(\left({\widehat{\overline{\mathbf{s}}}}^{\left(\mathrm{b}\right)}\odot {\boldsymbol{\upvarphi}}_n\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{t}}_0^{\left(\mathrm{b}\right)}\right)\right)\otimes {\mathbf{a}}_n\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{\boldsymbol{\upmu}}}_n^{\left(\mathrm{b}\right)}\right)\right)}^{\mathrm{H}}{\overline{\mathbf{x}}}_n}{{\left\Vert {\mathbf{a}}_n\left({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{\boldsymbol{\upmu}}}_n^{\left(\mathrm{b}\right)}\right)\right\Vert}_2^2\cdot {\left\Vert {\widehat{\overline{\mathbf{s}}}}^{\left(\mathrm{b}\right)}\odot {\boldsymbol{\upvarphi}}_n\Big({\widehat{\mathbf{p}}}^{\left(\mathrm{b}\right)},{\widehat{t}}_0^{\left(\mathrm{b}\right)}\Big)\right\Vert}_2^2} \) | M(2K + 1) + K |