Stochastic adaptive sensor modeling and data fusion

Abstract

One of the major problems in sensor fusion is that sensors frequently provide spurious observations which are difficult to predict and model. The spurious data from sensors must be identified and eliminated from the sensor fusion since its incorporation in the fusion pool might lead to inaccurate estimation. This paper presents a sensor fusion strategy based on Bayesian approach that can automatically identify the inconsistency in sensor data so that the spurious sensor data can be eliminated from the sensor fusion process. The proposed method adds a term to the commonly used Bayesian technique that represents the probabilistic estimate corresponding to the event that the data is not spurious conditioned upon the data and the true state. This term has the effect of increasing the variance of the posterior distribution when data from one of the sensors is inconsistent with respect to the other. The proposed strategy was verified with the help of extensive computations performed on simulated data from three sensors. A comparison was made between two different fusion schemes: centralized fusion in which data from all sensors were fused at once, and decentralized or sequential Bayesian method which provided opportunity to identify and eliminate spurious data from the fusion process. The simulations verified that the proposed strategy was able to identify spurious data, and its elimination from the fusion process led to better estimation.

Multi-source Information Fusion Technology and Its Engineering Application

“…The second is a random adaptive sensor model (Kumar, Garg, & Zachery, 2006), which detects data without using prior knowledge. It is developed in the Bayesian fusion framework.…”

Distributed Multisensor Data Fusion under Unknown Correlation and Data Inconsistency

“…In other words, the covariance provides a good approximation of all disturbances affecting the sensor measurements. However, in reality, uncertainties in sensor measurements may not only come from noise but also from unexpected situations, such as short duration spike faults, sensor glitches, permanent failure or slowly developing failure due to sensor elements [ 40 , 41 , 42 ]. Since these types of uncertainties are not attributable to the inherent noise, they are difficult to model.…”

“…However, sensors may produce inconsistent and spurious data due to unmodeled faults, including permanent sensor failures, sensor glitches, short duration spike faults, slowly developing failures due to sensor elements, etc. [ 40 , 41 , 42 ]. Fusion of inconsistent sensor data with correct data can lead to severely inaccurate results [ 43 ].…”

Multi-Sensor Data Fusion in Presence of Uncertainty and Inconsistency in Data

“…Sensor modeling (Manyika & Durrant-Whyte, 1994, Kumar et al, 2005b, Kumar et al, 2006a deals with developing an understanding of the nature of measurements provided by the sensor, the limitations of the sensor, and probabilistic understanding of the sensor performance in terms of the uncertainties. The information supplied by a sensor is usually modeled as a mean about a true value, with uncertainty due to noise represented by a variance that depends on both the measured quantities themselves and the operational parameters of the sensor.…”