Nonlinear Filtering With Sporadically Missing Sensor Data

The traditional Gaussian filtering adjustment is restricted with the assumption that the sensor data are received at every sampling instant. In practice, however, the observed data are often missing; thereupon, the estimates provided by the Gaussian filters are more likely unreliable. This letter addresses the irregularity of partially missing sensor data and, subsequently, designs an advanced Gaussian filter to tackle this irregularity. First, it proposes a modified measurement model that stochastically incorporates the partially missing data phenomenon. Subsequently, it reformulates the relevant parameters in reference to the modified measurement model. The newly derived parameters substitute the respective ones in the traditional Gaussian filtering and the proposed filtering method ensues. The simulation results obtained for two numerical examples conclude an improved filtering accuracy of the proposed filter in the presence of partially missing sensor data.