The Institution of Engineering and Technology Seminar on Target Tracking and Data Fusion

09.35 Keynote: Estimation without Independence Filtering and estimation algorithms assume conditional independence, either in the process or observation noises, however, in many systems such as those engaged in distributed data fusion, this assumption is incorrect. Robust filtering and estimation algorithms need to be developed which can automatically account for the unknown joint distribution. This talk will examine: • a set of algorithms in development which use weighted geometric means to replace the Bayes rule • the properties of these forms • examples illustrating the results Simon Julier, Senior Lecturer, UCL, UK