An Adaptive Kalman Fusion Technique for Reference Tracking Under Shot Noise

Sensor fusion enhances robotic perception by integrating data from multiple sensing modalities. This paper introduces a biologically inspired, closed-loop sensor fusion framework combining camera and LiDAR data to robustly estimate the position of a moving object. To mitigate non-Gaussian disturbances such as shot noise, the approach employs dynamic weighting strategies integrated with the Maximum Correntropy Criterion Kalman Filter (MCC-KF). The proposed method is validated experimentally using a custom-built platform, demonstrating enhanced robustness and accuracy under diverse and challenging noise conditions.