Abstract
Usually, the global navigation satellite system (GNSS)-based attitude determination and the GNSS-based relative positioning are treated as separate problems. In this contribution, the two problems are resolved as a combined solution of a joint model with multivariate constraints, in which the known geometry of the antenna array on the rover platform is used as constraints, thereby improving the parameter estimation in not only the attitude determination but also the relative positioning. The objective function of the joint model is decomposed and then reduced so that only the terms concerning the constraints are retained, thus transforming the objective function into the exact form of that of the GNSS-based multivariate constrained attitude determination model. Then the multivariate constrained least-squares ambiguity decorrelation adjustment method together with the efficient shrinkage or expansion with bounds strategies, which are shown to work well in the attitude determination, can be extended to the joint model. In experimental validation, the joint model is tested with both simulated and real GPS/BDS datasets in terms of ambiguity resolution success rate and the precision of solutions. The results confirm that the joint model ensures the highest ambiguity resolution success rate among the existing unaided GNSS-based relative positioning methods, and the improvement is more obvious when the observation model is weaker, for example, when the number of available satellites is fewer or when the precision of observations is poorer, thus showing an overall best performance in both reliability and precision aspects.
Similar content being viewed by others
References
Buist PJ, Teunissen PJG, Giorgi G, Verhagen S (2009) Multiplatform instantaneous GNSS ambiguity resolution for triple- and quadruple-antenna configurations with constraints. Int J Navig Observ 2009:1–14
Buist PJ, Teunissen PJG, Giorgi G, Verhagen S (2009b) Instantaneous GNSS-based kinematic relative positioning and attitude determination using multi-antenna configurations. In: Proceedings of the international symposium on GPS/GNSS 2009, Jeju, Korea, January, 1–12
Chen B (2005) Optimization theory and algorithm, 2nd edn. Tsinghua University, Beijing
Chen Y (2010) A new multipliers algorithm. Ph.D. thesis, Tongji University, Shanghai, China
Giorgi G, Teunissen PJG, Verhagen S, Buist PJ (2012) Instantaneous ambiguity resolution in global-navigation-satellite-system-based attitude determination applications: a multivariate constrained approach. J Guid Control Dyn 35(1):51–67
He H, Li J, Yang Y, Xu J, Guo H, Wang A (2014) Performance assessment of single- and dual-frequency BeiDou/GPS single-epoch kinematic positioning. GPS Solut 18(3):393–403
Hestenes MR (1969) Multiplier and gradient methods. J Optim Theory Appl 4(5):303–320
Pan S, Meng X, Gao W, Wang S, Dodson A (2014) A new approach for optimising GNSS positioning performance in harsh observation environments. J Navig 67(6):1029–1048
Powell MJD (1969) A method for nonlinear constraints in minimization problems. Optimization 5(6):283–298
Teunissen PJG (2007) The LAMBDA method for the GNSS compass. Artif Satell 41(3):89–103
Teunissen PJG (2010) Integer least-squares theory for the GNSS compass. J Geod 84(7):433–447
Teunissen PJG (2012) A-PPP: array-aided precise point positioning with global navigation satellite systems. IEEE Trans Signal Process 60(6):2870–2881
Teunissen PJG, de Jonge PJ, Tiberius CCJM (1997) The least-squares ambiguity decorrelation adjustment: its performance on short GPS baselines and short observation spans. J Geod 71(10):589–602
Teunissen PJG, Giorgi G, Buist PJ (2011) Testing of a new single-frequency GNSS carrier phase attitude determination method: land, ship and aircraft experiments. GPS Solut 15(1):15–28
Acknowledgments
This work is supported by the National Natural Science Foundation of China (Grant No. 41904014).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Proof of (12)
Appendix: Proof of (12)
With \({\varvec{Q}}_{{\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)}}^{ - 1} = \overline{\varvec{B}}^{\rm T} {\varvec{Q}}_{{\overline{\varvec{y}}\overline{\varvec{y}}}}^{ - 1} \overline{\varvec{B}}\), \(\overline{\varvec{B}} = \left[ {\begin{array}{*{20}c} {{\varvec{F}}^{\rm T} } & 0 \\ 0 & 1 \\ \end{array} } \right] \otimes {\varvec{G}}\), and \({\varvec{Q}}_{{\overline{\varvec{y}}\overline{\varvec{y}}}} = {\varvec{P}}_{n} \otimes {\varvec{Q}}_{{{\varvec{yy}}}}\), the second term of (11) equals
Let the LDLT decomposition of \({\varvec{P}}_{n}^{ - 1} { = }\left( {\frac{1}{2}{\varvec{D}}_{{\text{A}}} {\varvec{D}}_{{\text{A}}}^{\rm T} } \right)^{ - 1}\) be given as \({\varvec{P}}_{n}^{ - 1} = {\varvec{L}}_{n}^{{ - {\rm T}}} {\varvec{D}}_{n}^{ - 1} {\varvec{L}}_{n}^{ - 1}\), where \({\varvec{L}}_{n}^{ - 1} = \left[ {\begin{array}{*{20}c} 1 & {} & {} & {} & {} \\ \vdots & \ddots & {} & {} & {} \\ { - \frac{1}{i}} & { - \frac{1}{i}} & 1 & {} & {} \\ \vdots & \vdots & \vdots & \ddots & {} \\ { - \frac{1}{n}} & { - \frac{1}{n}} & { - \frac{1}{n}} & \cdots & 1 \\ \end{array} } \right]\) with i being the row number, and the diagonal entries of \({\varvec{D}}_{n}^{ - 1}\) equal to \(\frac{i + 1}{{2i}}\), \(i = 1,\;\ldots ,\;n\). If we set \({\varvec{M}} = - \frac{1}{n}{\varvec{e}}_{n - 1}^{\rm T}\), \({\varvec{P}}_{n}^{{{ - }1}}\) is rewritten as \({\varvec{P}}_{n}^{{{ - }1}} = \left[ {\begin{array}{*{20}c} {{\varvec{L}}_{n - 1}^{ - 1} } & {} \\ {{\varvec{M}}^{\rm T} } & 1 \\ \end{array} } \right]^{\rm T} \left[ {\begin{array}{*{20}c} {{\varvec{D}}_{n - 1}^{ - 1} } & {} \\ {} & {\frac{2n}{{n + 1}}} \\ \end{array} } \right]\left[ {\begin{array}{*{20}c} {{\varvec{L}}_{n - 1}^{ - 1} } & {} \\ {{\varvec{M}}^{\rm T} } & 1 \\ \end{array} } \right]\). Then it follows that
where \(\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right) = \hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}}} \right) + \left( {{\varvec{M}}^{\rm T} {\varvec{F}}^{\rm T} \otimes {\varvec{I}}_{3} } \right)\left( {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right)\) and \({\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right)\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right)}} = \frac{n + 1}{{2n}}{\varvec{Q}}_{{\hat{\varvec{b}}\left( {\varvec{a}} \right)\hat{\varvec{b}}\left( {\varvec{a}} \right)}}\) denote the conditional baseline solution of \({\varvec{b}}_{{{\text{m}}_{1} {\text{r}}}}\) (conditioned on \(\overline{\varvec{a}}\) and \({\varvec{X}}_{{\varvec{C}}}\)) and its v–c matrix, respectively. Since
with the property of the inverse matrix \(\left[ {\begin{array}{*{20}c} {\varvec{A}} & {\varvec{U}} \\ {\varvec{V}} & {\varvec{D}} \\ \end{array} } \right]^{ - 1} = \left[ {\begin{array}{*{20}c} {\left( {{\varvec{A}} - {\varvec{UD}}^{ - 1} {\varvec{V}}} \right)^{ - 1} } & { - \left( {{\varvec{A}} - {\varvec{UD}}^{ - 1} {\varvec{V}}} \right)^{ - 1} {\varvec{UD}}^{ - 1} } \\ { - \left( {{\varvec{D}} - {\varvec{VA}}^{ - 1} {\varvec{U}}} \right)^{ - 1} {\varvec{VA}}^{ - 1} } & {\left( {{\varvec{D}} - {\varvec{VA}}^{ - 1} {\varvec{U}}} \right)^{ - 1} } \\ \end{array} } \right]\), this v–c matrix is reformulated as \({\varvec{Q}}_{{\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)}} = \left[ {\begin{array}{*{20}c} {\left( {{\varvec{FP}}_{n - 1}^{ - 1} {\varvec{F}}^{\rm T} } \right)^{ - 1} } & { - \left( {{\varvec{FP}}_{n - 1}^{ - 1} {\varvec{F}}^{\rm T} } \right)^{ - 1} {\varvec{FM}}} \\ { - {\varvec{M}}^{\rm T} {\varvec{F}}^{\rm T} \left( {{\varvec{FP}}_{n - 1}^{ - 1} {\varvec{F}}^{\rm T} } \right)^{ - 1} } & {\frac{n + 1}{{2n}} + {\varvec{M}}^{\rm T} {\varvec{F}}^{\rm T} \left( {{\varvec{FP}}_{n - 1}^{ - 1} {\varvec{F}}^{\rm T} } \right)^{ - 1} {\varvec{FM}}} \\ \end{array} } \right] \otimes {\varvec{Q}}_{{\hat{\varvec{b}}\left( {\varvec{a}} \right)\hat{\varvec{b}}\left( {\varvec{a}} \right)}}\).
Thus, we get
Substitution of (20) into the last equation of (19) gives
\(\begin{aligned} \left\| {\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right) - \overline{\varvec{b}}} \right\|_{{{\varvec{Q}}_{{\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}\left( {\overline{\varvec{a}}} \right)}} }}^{2} & = \left( {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right)^{\rm T} {\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)}}^{ - 1} \left( {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right) + \left[ {\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right) - {\varvec{b}}_{{\text{m}_{1} \text{r}}} } \right]^{\rm T} {\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right)\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right)}}^{ - 1} \left[ {\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right) - {\varvec{b}}_{{\text{m}_{1} \text{r}}} } \right] \\ & = \left\| {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right\|_{{{\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)}} }}^{2} + \left\| {\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},{\varvec{X}}_{{\varvec{C}}} } \right) - {\varvec{b}}_{{\text{m}_{1} \text{r}}} } \right\|_{{{\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},{\varvec{X}}_{{\varvec{C}}} } \right)\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},{\varvec{X}}_{{\varvec{C}}} } \right)}} }}^{2} \\ \end{aligned}\) in which \(\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}},X_{{\varvec{C}}} } \right) = \hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}}} \right) + \left( {{\varvec{M}}^{\rm T} {\varvec{F}}^{\rm T} \otimes {\varvec{I}}_{3} } \right)\left( {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right) = \hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}}} \right) - {\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{2} \left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)}} {\varvec{Q}}_{{\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right)}}^{ - 1} \left( {\hat{\overline{\varvec{b}}}_{1} \left( {\overline{\varvec{a}}} \right) - {\varvec{X}}_{{\varvec{C}}} } \right)\).
□
Rights and permissions
About this article
Cite this article
Wu, S., Zhao, X., Pang, C. et al. Improving ambiguity resolution success rate in the joint solution of GNSS-based attitude determination and relative positioning with multivariate constraints. GPS Solut 24, 31 (2020). https://doi.org/10.1007/s10291-019-0943-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10291-019-0943-y