Spatiotemporal Bias Compensation And Data Fusion For Multisensor Systems

Multisensor data fusion has been widely applied in various fields due to its ability to use multisource information to improve the overall system performance.A significant prerequisite for effective fusion is to accurately estimate and compensate the system biases of multisensor.The eixsting works mainly focus on sensor spatial biases,while there also exists temporal biases between sensors in practical applications.The eixsting methods cannot be applied to general asynchronous multisensor systems that contain both spatial and temporal biases,and there are many problems to be solved: 1)The existing methods lack effective strategy to handle the measurements of asynchronous multisensor with spatiotemporal biases.Accurate filtering models in the presence of spatiotemporal biases need to be formulated,so that simultaneous estimation of target state and spatiotemporal bias can be achieved.2)The target may maneuver in practical applications,and there is still a lack of research that can deal with the spatiotemporal biases and maneuvering target tracking in asynchronous multisensor systems.3)In multitarget scenarios,the measurements of multitarget all contain useful information about sensor spatiotemporal biases.It is still an open problem to effectively utilize bias information in multitarget measurements to further improve the system performance.4)Measurement origin uncertainty is another important factor that affects data fusion,and it is highly coupled with spatiotemporal biases.The data fusion in the presence of both spatiotemporal biases and measurement origin uncertainty has not been properly handled by any existing methods.In order to cope with the mentioned problems,this dissertation is dedicated to the research on spatiotemporal bias compensation and data fusion.The main contributions of the dissertation are as follows:1.To deal with the problem of spatiotemporal biases in practical asynchronous multisensor systems,a batch processing and a sequential processing simultaneous spatiotemporal bias and target state estimation methods are proposed.The general situation that sensors sample at different times with different and varying periods is explored,and unknown time delays may exist between time stamps and true measurement times.Due to the time delays,the true measurement interval of the measurements from different sensors is available and may be different from their time stamp interval.The unknown difference between the time delays of different sensors is considered as the temporal bias,which is augmented into the state vector to be estimated with along the spatial bias.The idea is to use the time stamp interval in state transition and formulate the feasible augmented state equation.Multisensor measurements are collected in batch processing or sequential processing schemes.It is found the time of transited target state is unequal to the measurement time,and their difference equals the temporal bias.The temporal bias is then used to align the target state with the measurement,and the measurement is formulated as functions of both target states and spatiotemporal biases in the two processing schemes.Since both the biases and target state are part of the augmented state,we can estimate the bias and target state under the same filtering framework.Simulation results demonstrate that the proposed methods can achieve accurate estiamtion of both spatiotemporal bias and target state under different bias values and measurement noise levels.The proposed methods can also realize bias compensation and data fusion under the test of real data.2.To deal with the problem of maneuvering target tracking in asynchronous multisensor systems with spatiotemporal bias,a simultaneous spatiotemporal bias and maneuvering target state estimation method is proposed.The augmented state equations are presented to formulate the nearly coordinated turn mo tion and the nearly constant acceleration motion,respectively,without exactly known measurement interval.In each target motion,the measurements are formulated as functions of both spatiotemporal bias and target state based on the time difference between the measurements and the states to be estimated.Furthermore,the interacting multiple model estimator is incorporated with the unscented Kalman filter to achieve simultaneous sequential estimation of spatiotemporal biases and target states in the presence of target maneuvers.Simulation results demonstrate that the proposed method can achieve accurate maneuvering target tracking in the presence of spatiotemporal bias.3.To deal with the problem of effectively utilizing the spatiotemporal bias information in multisensor multitarget measurements,an sequential spatiotemporal bias compensation and fusion method based on the minimum mean square error framework is proposed.When updating the augmented state estimates of a target,the spatiotemporal bias estimates of the previous target are used as linear pseudo-measurements,which are combined with the sensor measurements to serve as the augmented measurements for the target.The correlation between the pseudo-measurement and the augmented state estimates of this target from the previous time step is analyzed.To handle the correlation problem,a novel estimator is derived under the minimum mean square error framework,which can update both target state and spatiotemporal bias estimates sequentially using the measurement of each target.This can avoid high-dimensional processing when augmenting all target measurements,and has the advantage of being efficient and flexible.Simulation results demonstrate that the proposed method can simultaneously estimate the spatiotemporal bias and each target state,and the spatiotemporal bias estimates are sequentially improved along with the filtering process of each target.4.To deal with data fusion problem in the presence of both spatiotemporal bias and measurement origin uncertainty,a spatiotemporal bias compensation,multitarget association and fusion method is proposed for asynchronous multisensor multitarget tracking systems.The spatiotemporal biases of the sensors are considered as a part of the state vector to be estimated,and the state space model in the presence of spatiotemporal bias is formulated.Based on the formulation,the joint data association and bias compensation problem is converted into the classical data association and filtering problem in a unified Bayesian framework,wi thout requirement of iterative optimization procedure between data association and bias estimation.The data association is formulated as a 2-D assignment problem and is solved using the generalized Auction algorithm.In this method,the measurements reported by each sensor are associated sequentially with the tracks.For each track-measurement pair in the association results,the unscented Kalman filter is used to handle the measurement to produce estimates of target states and spatiotemporal bias simultaneously.Furthermore,a bias fusion approach with feedback is presented to fuse the spatiotemporal bias estimates from each track to improve the bias compensation performance and solve the correlation problems therein.Simulation results demonstrate that the proposed method can achieve good association,compensation and fusion performance under different detection probabilities,temporal bias values,and in the case with varying target number and intersection tracks.