Multisource Interactive Resilient Fusion Algorithm Based on RIEKF

As the number of heterogeneous redundant sensors on remotely piloted vehicle (RPV) increases, onboard sensors require a more rational and efficient credibility evaluation system and a resilient fusion framework to achieve the essence of seamless sensor group switching. A simple and efficient sensor credibility evaluation system is proposed to guide the selection of the optimal multisource sensor submodel combination, thereby providing key model prior knowledge for multisource resilient fusion. Furthermore, a multimodel interactive resilient fusion framework based on right invariant extend Kalman filter (RIEKF) is proposed, utilizing the defined sensor credibility indexes to guide the design of the model transition probability matrix, thereby reducing the sensitivity of submodel weights to fusion stability and solving the problem of the model transition matrix lacking a basis for adjustment. Model weights are updated in real time through credibility prior information and submodel posterior probabilities, thus leveraging the adaptive resilience advantage between models to achieve seamless switching between submodels in complex environments. Experimental results show that the algorithm presented in this article, without using any sensor fault diagnosis and isolation logic, and without setting any complex detection timing and thresholds, demonstrates a resilience advantage, thereby enhancing the adaptability of the state estimation system in complex environments.