Harris-Lovassy N-Dimensional Rule of Combination
Daniel Harris, Peter Lovassy, Darin T. Dunham · 2022 IEEE Aerospace Conference (AERO) · 2022
This paper presents a novel implementation of the principle contained in the Dubois- Prade rule of combination (DP- RoC) in order to overcome the non-associative property that exists in its original implementation. Various data fusion rules of combination have been presented in the last 30 years to deal with the uncertainty distribution associated with combining different and conflicting data sources (e.g., multi-sensor measurements), such that when combined should not have equal uncertainties assigned to the probabilistic beliefs. This new implementation is called the Harris- Lovassy N-Dimensional Rule of Combination (HL-NRoC), and like DP-RoC, it assumes the principle that given conflict between two (or more) sources, at least one of the sources is correct, however this implementation overcomes the non-associative property by achieving permutation symmetry via an n-ary operator. Implementing this principle in a manner that achieves permutation symmetry is key to overcoming the counter-intuitive results that may result from fusing conflicting overconfident sources. Testing was conducted and presented with conflicting and overconfident sources which demonstrates the value of HL-NRoC by achieving similar performance to other commonly used methods, but with better calibrated probability values - by as much as 90%.