Asynchronous Fusion Algorithm under Correlated Noise Conditions

In complex real-time monitoring and control systems, multi-sensor data fusion technology plays a crucial role. Due to factors such as sensor characteristics and communication delays, synchronized data acquisition is often difficult to achieve, making asynchronous data fusion a more practical research direction that better aligns with real-world requirements.

To address asynchronous multi-sensor data fusion problems in continuous-time systems, researchers have developed an effective processing solution. The implementation begins with performing asynchronous discretization at both the fusion center and individual sensor nodes. This critical step converts the continuous-time system into a discrete-time system, facilitating subsequent mathematical modeling and algorithm implementation. Notably, the discretized system exhibits unique noise characteristics: although the process noise and measurement noise in the original continuous-time system, along with measurement noises from different sensors, are mutually uncorrelated, these noises may become correlated after discretization.

To handle challenges arising from this correlated noise, researchers have derived specialized asynchronous fusion algorithms. The algorithm implementation employs covariance intersection techniques and handles cross-correlation terms through modified Kalman filtering approaches. It fully accounts for correlations between noises after discretization, ensuring the accuracy and reliability of fusion results. By establishing appropriate mathematical models and incorporating advanced filtering techniques, the algorithm can effectively integrate multi-source information under asynchronous data arrival conditions, thereby enhancing the system's overall perception and decision-making capabilities. Key functions include timestamp synchronization handlers and correlation-aware prediction-update cycles.

Simulation experiments have validated the algorithm's effectiveness. In simulated asynchronous multi-sensor environments, the algorithm maintains stable performance even under complex correlated noise conditions, significantly improving fusion accuracy. This holds substantial reference value for practical applications such as autonomous driving, industrial monitoring, and aerospace fields, providing a feasible technical pathway for solving noise-related problems in asynchronous data fusion. Code implementation typically involves queue management for out-of-sequence measurements and adaptive covariance matrix adjustments.