Conditionally-Minimax Nonlinear Filtering for Continuous-Discrete Stochastic Observation Systems: Comparative Study in Target Tracking

Авторы: Borisov Andrey V., Bosov Alexey V., Miller Gregory B.
Сборник: 2019 IEEE 58th Conference on Decision and Control (CDC)
Год издания: 2019
Место издания: IEEE
Первая страница: 2586
Последняя страница: 2591
DOI: 10.1109/cdc40024.2019.9029876
Аннотация: The paper presents an algorithm of the Conditionally-Minimax Nonlinear Filtering (CMNF) of the stochastic differential system state given indirect noisy discrete-time observations. The filter has a two-step "prediction-correction" form with fixed basic prediction and correction terms, which are jointly processed in the optimal way. The obtained estimates are unbiased and the theoretical covariance matrix of CMNF errors, which may be calculated a priori, meets the real a posteriori values. The key feature of the CMNF algorithm is a possibility to choose the preliminary observation transformation, basic prediction and correction functions for a particular case of the observation system, which allows to improve the estimation accuracy significantly. All the features of CMNF are demonstrated by the maneuvering target tracking model in comparison with the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) estimates.
Добавил в систему: Борисов Андрей Владимирович

	ИСТИНА	Войти в систему Регистрация
	ИПМех РАН
	Главная Поиск Статистика О проекте Помощь

ИСТИНА