Evidence supporting measure of similarity for reducing the complexity in information fusion☆

Abstract

This paper proposes a new solution for reducing the number of sources of evidence to be combined in order to diminish the complexity of the fusion process required in some applications where the real-time constraint and strong computing resource limitation are of prime importance. The basic idea consists in selecting, among the whole set of sources of evidence, only the biggest subset of sources which are not too contradicting based on a criterion of Evidence Supporting Measure of Similarity (ESMS) in order to process solely the coherent information received. The ESMS criterion serves actually as a generic tool for outlier source identification and rejection. Since the ESMS between several belief functions can be defined using several distance measures, we browse the most common ones in this paper and we describe in detail the principle of our Generalized Fusion Machine (GFM). The last part of the paper shows the improvement of the performances of this new approach with respect to the classical one in a real-data based and real-time experiment for robot perception using sonar sensors.

Evidence Combination From an Evolutionary Game Theory Perspective

“…Recently, different fields have expressed concerns over the applications of DST, such as parameter estimation [36], [37], classification and clustering [38]–[40], decision-making [41]–[44], etc [45]–[51]. And, as an extension of this theory, Dezert-Smarandache Theory (DSmT) of plausible and paradoxical reasoning [52]–[54] has also been given much attention [27], [55]. A few basic concepts on DST are introduced below.…”

A method of evidence of combination based on the improved similarity measure

“…Presently, many distance measure, i.e. Pignistic distance [11], [24], Cosine distance [25], [26], [27], and Jousselme distance [25], are widely used. However, these distance measures is to describe and characterize the degree of conflict according to the corresponding focal elements between two evidences, where the sequence of the size of each assignment is not considered.…”

The problem of entropy production in the classic rule of combination in the Dezert-Smarandache theory

“…Many experts and scholars made great achievements in this field [1]- [13]. In the Shafer model, an ultimate refinement of the problem was possible so that singleton focal elements were supposed to be exclusive and exhaustive in the closed-world.…”

“…In the Shafer model, an ultimate refinement of the problem was possible so that singleton focal elements were supposed to be exclusive and exhaustive in the closed-world. Therefore, Dempster [6] proposed a wellknown combinational rule on the basis of the Shafer model in (1) according to the Dempster-Shafer theory (DST). This rule was widely applied in different fields.…”