[Paper Review] An In-Depth Look at Information Fusion Rules & the Unification of Fusion Theories

Florentín Smarandache|arXiv (Cornell University)|Oct 14, 2004

Multi-Criteria Decision MakingDecision Sciences19 references16 citations

TL;DR

This paper presents a comprehensive survey and unification of 32 information fusion rules, introducing new rules like PCR5 for exact conflict redistribution and proposing a Unification of Fusion Theories (UFT) that extends Boolean algebra over the frame of discernment to handle epistemic uncertainty and paradoxical/conflicting evidence. The framework selects the most appropriate rule per application context, with DSmT as the default theory and PCR5 as the most mathematically precise conflict redistribution method.

ABSTRACT

This paper may look like a glossary of the fusion rules and we also introduce new ones presenting their formulas and examples: Conjunctive, Disjunctive, Exclusive Disjunctive, Mixed Conjunctive-Disjunctive rules, Conditional rule, Dempster's, Yager's, Smets' TBM rule, Dubois-Prade's, Dezert-Smarandache classical and hybrid rules, Murphy's average rule, Inagaki-Lefevre-Colot-Vannoorenberghe Unified Combination rules [and, as particular cases: Iganaki's parameterized rule, Weighting Average Operator, minC (M. Daniel), and newly Proportional Conflict Redistribution rules (Smarandache-Dezert) among which PCR5 is the most exact way of redistribution of the conflicting mass to non-empty sets following the path of the conjunctive rule], Zhang's Center Combination rule, Convolutive x-Averaging, Consensus Operator (Josang), Cautious Rule (Smets), ?-junctions rules (Smets), etc. and three new T-norm & T-conorm rules adjusted from fuzzy and neutrosophic sets to information fusion (Tchamova-Smarandache). Introducing the degree of union and degree of inclusion with respect to the cardinal of sets not with the fuzzy set point of view, besides that of intersection, many fusion rules can be improved. There are corner cases where each rule might have difficulties working or may not get an expected result.

Motivation & Objective

• To systematize and extend existing information fusion rules, including classical and novel approaches, for handling epistemic uncertainty and conflicting evidence.
• To address the limitations of individual fusion rules, which fail under specific conditions such as high conflict or incomplete information.
• To propose a Unification of Fusion Theories (UFT) that dynamically selects the optimal fusion model, rule, and algorithm based on application context.
• To improve fusion accuracy by introducing degree of union and inclusion metrics based on set cardinality, not fuzzy membership.
• To formalize a Boolean algebraic structure over the frame of discernment, incorporating complementation and closure under union, intersection, and complement.

Proposed method

• Extends fusion rules from the power set 2Θ to the super-power set SΘ = {φ, θ1, θ2, ..., C(θ1 ∪ θ2)} to include complements and handle non-exclusive elements.
• Introduces new fusion rules such as Proportional Conflict Redistribution (PCR5), which redistributes conflicting mass exactly along the path of the conjunctive rule.
• Applies T-norm and T-conorm operations derived from neutrosophic and fuzzy set theory to enhance fusion under uncertainty.
• Proposes a decision-based selection mechanism in UFT, where the most suitable rule is chosen per scenario (e.g., conjunctive for reliable sources, disjunctive for optimistic fusion).
• Uses the conjunctive rule as a foundation, followed by conflict mass redistribution via rules like PCR5, Iganaki’s parameterized rule, or weighted averaging.
• Introduces quasi-associative and quasi-Markovian algorithms to preserve associativity and Markovian behavior in dynamic, real-time fusion systems.

Experimental results

Research questions

• RQ1 Which fusion rule performs optimally under conditions of high conflict, incomplete information, or unreliable sources?
• RQ2 How can conflicting mass be redistributed in a mathematically exact way that preserves the logical path of the conjunctive rule?
• RQ3 What algebraic structure best supports the unification of diverse fusion theories while handling non-exclusive, fuzzy, or neutrosophic elements?
• RQ4 Can the degree of union and inclusion—based on set cardinality—improve fusion rule performance beyond traditional fuzzy or probabilistic measures?
• RQ5 How can a dynamic, application-specific fusion framework be designed to select the optimal rule, model, and algorithm in real time?

Key findings

• PCR5 is identified as the most mathematically exact method for redistributing conflicting mass, following the conjunctive rule’s path precisely.
• The conjunctive rule fails when any belief mass is zero, highlighting the need for conflict redistribution mechanisms.
• The disjunctive and exclusive disjunctive rules also suffer from zero-mass propagation, limiting their robustness in uncertain environments.
• UFT provides a flexible, application-driven framework where the default theory is DSmT, and the default rule is the average minimum principle.
• The super-power set SΘ, closed under union, intersection, and complement, forms a Boolean algebra that enables richer fusion modeling than traditional power sets.
• Normalization is not required for all rules; incomplete or paraconsistent masses can be fused and normalized post-combination or left unnormalized depending on context.