Hybrid state estimation and model-set design of invariable-structure semi-ballistic reentry vehicle

YongQi Liang, ChongZhao Han
2011 Science China Information Sciences  
Multiple-model approach is one of the main streams for hybrid estimation. The difficulty of this approach to estimate the hybrid state of the semi-ballistic reentry vehicle (SBRV) is model-set design. This paper proposes a quasi-Monte Carlo model set that can ensure the estimator near-optimal in the sense of minimum mean square error (MMSE). The SBRV has a high nonlinearity and its mode is spanned by multiple parameters with known bounds. The design methods and characteristics of the
more » ... Carlo model set are given. The proposed model set has a higher accuracy than the model-set generated by the Monte Carlo method. The theoretical analysis and simulation results show the effectiveness and reasonability of the newly designed model set. Keywords semi-ballistic reentry vehicle, hybrid estimation, model-set design, Monte Carlo method, quasi-Monte Carlo method Citation Liang Y Q, Han C Z. Hybrid state estimation and model-set design of invariable-structure semi-ballistic reentry vehicle. 813 possible modes, and its output is the combination of each model-based output. The performance of the MM approach depends largely on the set of models used, and model-set design is one of the most important tasks in the applications of the MM approach [13] . In this paper, we mainly discuss the model-set design for the MM-SBRV estimator. Model-set design is not only important but also difficult. Its design is still largely in the kingdom of black magic or the domain of art [13] . A model set has a fixed or variable structure. The former is most widely used, which is applied to the autonomous multiple-model (AMM) approach or the cooperative multiple-model (CMM) approach [11, 14, 15] . Each model of the AMM estimator works individually and independently. The advantage of the AMM estimator over the non-MM estimator stems from its superior processing of the outputs of all these models to generate the overall estimate [11] . In this paper, we use the AMM approach to estimate the hybrid state of the invariable-structure SBRV. Magill [14] used a model set that is equal to the mode space for optimal estimation in the sense of minimum mean square error (MMSE). In fact, the mode space is usually much larger than the model set [16, 17] . In this case, there are few general model-set design methods. Given bounds of parameters, [18] proposed a model set design method for linear systems. Based on the known bounds of a parameter, [4] designed model sets in the one-dimensional mode space of a nonlinear system. With prior information of the distribution of the mode, Li et al. [17] presented a theoretical description of model-set design, where the true mode and the model are viewed as random variables, and then the random model is designed to approximate the distribution of the mode. Based on this description, Li et al. proposed three model-set design approaches. As a further development of [17], [19] designed model set for the multi-dimensional mode space in the minimum-mismatch sense. Although the MM approach was widely used for hybrid estimation, it is seldom used for the reentry problem. The difficulty lies in the design of the model set, and also none of the model-set design approaches mentioned above fits the characteristics of the SBRV. The vehicle has a high nonlinearity, and the model set in [18] cannot be used; its mode space is spanned by multi-parameters, then the model set in [4] cannot be used; its mode is easily known for bounds and its distribution is often unknown, thus the model set in [17] and [19] cannot be used. The model-set design approach proposed in this paper fits the characteristics of the SBRV, which is distributed in the mode space, easier to use, and ensures the estimator near-optimal in the sense of minimum mean square error (MMSE). This paper is organized as follows: The SBRV model with characteristics of its motion and mode are introduced in section 2. The near-optimal hybrid state estimator of the SBRV in the MMSE sense is designed in section 3. The quasi-Monte Carlo model set is proposed in section 4, along with its theoretical accuracy, design methods, and characteristics. Simulation results are provided in section 5. Conclusions are given in section 6.
doi:10.1007/s11432-010-4159-6 fatcat:fnuzil5imbd5dek2crqt7kinha