In this study, a new heuristic multiple model filter, called Multiple Model Extended Continuous Ant Colony Filter, is proposed to solve a nonlinear multiple model state estimation problem. In this filter, a bank of extended continuous ant colony filters are run in parallel to solve the multiple model estimation problem. The probability of each model is continually updated and consequently both the true model and the states of the nonlinear system are updated based on the weighted sum of the filters. The new multiple model filter is tested on an engineering problem. The problem is to estimate simultaneously the states of a fixed-wing unmanned aerial vehicle as well as the wind model, applied to the system. Four different wind models are considered and the proposed filter is unaware of the wind type. Then, observability of the states and the wind components are analyzed. Four new propositions are introduced and proved for unknown input observability, state and unknown input observability, the effect of time-varying unknown input matrix on the unknown input observability, and the effect of linearization errors on the state observability. Moreover, observability of the wind parameters is analyzed based on the nonlinear systems observability theory. Performance of the proposed filter is also evaluated in maneuvering flight and compared to a single extended continuous ant colony filter and a multiple model extended Kalman filter. A hardware-in-the-loop experiment is also performed to verify the real-time implementation capability of the suggested architecture.

在这项研究中，一种新的启发式多模型过滤器，称为多模型扩展连续蚁群过滤器提出了解决非线性多模型状态估计问题的方法。在此滤波器中，并行运行一排扩展的连续蚁群滤波器，以解决多模型估计问题。每个模型的概率不断更新，因此，真实模型和非线性系统的状态都将根据滤波器的加权和进行更新。新的多模型过滤器已针对工程问题进行了测试。问题是要同时估计应用于该系统的固定翼无人机的状态以及风模型。考虑了四种不同的风模型，并且所提出的过滤器不了解风类型。然后，分析状态和风分量的可观察性。引入并证明了四个新的命题未知的输入可观测，状态和未知输入观测性，随时间变化的未知输入矩阵的对未知输入可观测的效果，和线性化误差对状态观测性的效果。此外，基于非线性系统可观测性理论，分析了风参数的可观测性。拟议的过滤器的性能也在机动飞行中进行了评估，并与单个扩展的连续蚁群过滤器和多模型扩展的卡尔曼过滤器进行了比较。还执行了硬件在环实验，以验证所建议体系结构的实时实现能力。