A Sequential Dempster Shafer Particle Filter

I'm pleased to say that Johnny Worthy's work on extending Dempster Shafer evidential reasoning to sequential particle filters has been accepted as a journal article in IEEE Transactions on Aerospace and Electronic Systems. The paper, titled "Sequential Estimation from Uninformative Priors Using Dempster-Shafer Theory," has particular value when initiating sequential estimation processes with unobservable priors.

In the paper, he rigorously constructs belief mass functions used to update both the belief and plausibility of individual particle hypotheses. Using Dempster's rule of combination, the collective belief and plausibility 'surfaces' can be constructed. Interestingly, he is able to show that as the problem becomes observable, the belief and plausibility surfaces collapse on to one another and exactly match a Bayesian particle filter at that instant. A final advantage of the approach is that belief may be assigned to the hypothesis that the true state does not lie within the admissible region, thus letting the operator know if their admissible region assumptions are incorrect.

Dr. Worthy was able to validate the approach on real data from the GT-SORT telescope using observations of the tumbling COSMOS 1247 space object.

Please join me in congratulating Johnny on this accomplishment!

A link to the published paper will be added when the paper is posted onine.