多传感器噪声方差未知情况下的异步航迹融合

Asynchronous Track Fusion in Multi-Sensor Systems without Knowledge of Noise Covariance

针对分布式多传感器数据融合系统,提出了一种多传感器异步航迹融合算法。现有的多传感器信息融合算法大都基于Kalman滤波器,要求噪声方差已知,并且假定各传感器同步采样,不考虑通信延迟。本文在分布式处理的模式下,基于各传感器在扩展记忆因子递推最小平方(EFRLS)估计形成本地航迹的基础上,提出了一种融合误差均方差矩阵的迹最小意义下的异步目标航迹融合算法。仿真实验结果表明,这种融合算法是有效的,算法接近集中式融合算法的精度。

For distributed multisensor data fusion system, a multisensor asynchronous track fusion algorithm is proposed. In multi-sensor information fusion, the studied track fusion algorithm is based on the optimal Kalman filter,the filter requires knowledge of the noise covariance. And the present distributed estimation architectures assume that the sensors used are synchronous and no communication delays exist. But in the real environment, the noise variance is unknown in practice and the sensors used are asynchronous. An algorithm of asynchronous track fusion by minimizing the trace of the fusion error covariance matrix is presented from the extended forgetting factor recursive least square estimator. Simulation shows a satisfied precision in their performance, it approaches the centralized processing architecture.