1.3.3 - State Estimation Using Virtual Measurement Information
- Event
- 18. GMA/ITG-Fachtagung Sensoren und Messsysteme 2016
2016-05-10 - 2016-05-11
Nürnberg, Germany - Chapter
- 1.3 Modellbildung und Informationsfusion
- Author(s)
- B. Noack, U. Hanebeck - Karlsruhe Institute of Technology (KIT), Karlsruhe (Germany)
- Pages
- 72 - 77
- DOI
- 10.5162/sensoren2016/1.3.3
- ISBN
- 978-3-9816876-0-6
- Price
- free
Abstract
The computation of an estimate for the unknown state of a dynamical system is a central challenge in many disciplines and applications. In general, the estimation quality is directly tied to the amount of sensor data available to the state estimation system. However, insights from virtual or missing observations may also convey exploitable information on the system’s state. Such virtual measurement information may relate to constraints to which the state is subject. For instance, constraints to acceleration and turn rate of a mobile robot may apply and can be exploited. Analogously, missing observations that are attributable to obstacles can be translated into usable information, which is often referred to as negative sensor evidence. Such implicit information has to be reformulated into virtual measurement data in order to take advantage of it. As the Kalman filter and its derivatives are most widely used in state estimation applications, specific measurement and noise models for virtual observations are to be derived that can easily be integrated into the prediction-correction cycle of the Kalman filter. In this work, a set-membership representation of virtual measurement information is discussed.