Adaptive center error entropy STCKF by using fuzzy-BLS for UAV sensor data denoising

To address the issue of low denoising accuracy of unmanned aerial vehicle (UAV) sensor data in a nonlinear non-Gaussian system, an adaptive central error entropy (CEE)—strong tracking cubature Kalman filter (STCKF) algorithm based on fuzzy broad learning system (fuzzy-BLS) is proposed in this paper. Although entropy algorithms are known to be effective for denoising in non-Gaussian systems, their application in nonlinear systems is still limited. To address this issue, this study combines the central error entropy criterion with the STCKF algorithm. This approach is boosted by the denoising capabilities of the STCKF algorithm for nonlinear systems, thereby compensating for the shortcomings of the CEE criterion for nonlinear systems and leveraging the advantages of CEE in non-Gaussian systems. Thus, the new algorithm has enhanced robustness and accuracy for nonlinear non-Gaussian systems. To further optimize this algorithm, a parameter update method based on fuzzy-BLS is adopted to address the problem of excessive reliance on experience and lack of dependency in the selection of parameters, such as weight and kernel width, in the fusion of the CEE criterion. This method can dynamically adjust the optimal parameter template obtained from offline training online to minimize the root mean square error of the denoising results and provide adaptive denoising capability. Simulation and actual data denoising experiments confirmed that the proposed data denoising method accurately addresses the denoising problem of UAV sensor data in nonlinear non-Gaussian systems.