Elsevier

Information Fusion

Volume 27, January 2016, Pages 54-63
Information Fusion

An asynchronous sensor bias estimation algorithm utilizing targets’ positions only

https://doi.org/10.1016/j.inffus.2015.05.003Get rights and content

Highlights

  • A solution to the bias estimation problem for asynchronous sensors is proposed.

  • The B-spline interpolation method is used to realize time registration.

  • The derivation of the pseudo measurement only depends on target position.

  • The proposed algorithm is not sensitive to target’s maneuvre.

Abstract

Bias estimation is a critical problem in multi-sensor tracking systems, and most existing research has focused on the bias estimation of synchronous sensors; however, in practical applications, sensor measurements are usually asynchronous. The primary contribution of this paper is that a novel algorithm using B-spline interpolation time registration to achieve asynchronous sensor bias estimation is proposed. First, measurements are transformed into synchronous data using the B-spline interpolation time registration method. The time registration results are expressed as weighted results of the measurements. Second, a pseudo measurement equation is created based on the synchronous data. Compared with the pseudo measurements of other algorithms that use weighting coefficients, which are calculated by the target’s state, including the target’s velocity and time of arrival (TOA), a pseudo measurement that only depends on the target’s position can be derived. Thus, the problem of asynchronous sensor bias estimation, particularly with manoeuvring targets, can be solved effectively by the proposed algorithm. Finally, the effectiveness of the proposed algorithm is verified by simulations with the target performing s-shaped manoeuvres. Monte Carlo simulation results indicate that the Cramer-Rao lower bound (CRLB) is achievable; thus, the proposed algorithm is statistically efficient.

Introduction

Data fusion is an important technology in multi-sensor tracking systems that can improve tracking accuracy; however, without bias estimation, the integrated performance of data fusion will be poor [1]. Therefore, it is necessary to estimate sensors’ offset biases before data fusion. Thus, bias estimation is a critical problem in multi-sensor tracking systems.

To estimate the offset bias vector, two classical approaches are typically used: one is offline [2], and the other is online, which has better real-time performance. The online method was proposed by Nabaa and Bishop [3], who augmented the system state to include the bias vector as part of the state and then implemented an augmented state Kalman filter (ASKF) by stacking the states of all targets; the sensor biases were then recorded in a signal vector. The disadvantage of this method is that the dimension of the system vector may be very large, which can increase computational demand. Hsieh [4] improved this approach using a two-stage Kalman filter (KF), which could mitigate the complexity of the filtering process; however, this method could only be applied under specific conditions [5]. An alternative approach that is less computationally demanding is based on pseudo measurements under earth center earth fixed (ECEF) coordinates [6], [7], [8].

However, in practical applications, the target states detected by different sensors are not necessarily concurrently measured because of different sampling rates and the initial scanning times of the sensors. However, all of the above studies and other well-known works [9], [10], [11] have investigated only synchronous data; the bias estimation problem for asynchronous sensors [7], [12], [13], [14] has been studied considerably less.

In [7], the method transformed the measurements from different times of the sensors into pseudo measurements in “proper time slots,” creating pseudo measurements based on the measurements in the slot. Reference [13] extended the work of [7]: a pseudo measurement was obtained based on a current measurement and recent previous measurements from multiple sensors. In [14], the pseudo measurement equation of sensor biases was obtained according to all measurements in a fusion period. Reference [15] extended the work of [7] into 3D scenarios and concluded that the variance of bias estimates cannot be ignored when there are more uncertainties in bias estimation. However, the derivations of the pseudo measurements in [7], [12], [13], [14] were related to the transition matrix for CV motion, which is based on the assumption of a target with uniform motion; thus, those pseudo measurements were related to the target’s state including its velocity, and thus depended on the target state. As a result, that algorithm is sensitive to target motion.

The purpose of this paper is to propose a pseudo measurement that is only dependent on the target’s position. As a result, the merit of the proposed algorithm is insensitive to the target’s manoeuvres.

The remainder of this paper is organized as follows. The bias estimation method of synchronous sensors based on ECEF coordinates is reviewed in Section 2. The primary section of this paper discusses bias estimation of asynchronous sensors in Section 3. In Section 4, the effectiveness of the proposed method is demonstrated. Lastly, conclusions are presented in Section 5.

Section snippets

Observation model

Assuming that in the sensor’s local Cartesian coordinates, the true position of target is X1(k)=x1(k),y1(k),z1(k)T at time k, the corresponding measurements in the polar coordinate and local Cartesian coordinate are then r(k),θ(k),φ(k) and X1(k)=x1(k),y1(k),z1(k)T, respectively. The offset bias vector, which includes biases of range, azimuth and elevation, is ξ(k)=Δr(k),Δθ(k),Δφ(k)T. Transforming measurements into the local Cartesian coordinate, we can obtain the observation model:x1(k)=r-Δ

Bias estimation of asynchronous sensors

As shown in Section 2, the bias estimation of synchronous sensors can be efficiently solved by the method above. However, asynchronous sensors must still be studied. This section examines the derivation of the pseudo measurement based on the interpolation time registration algorithm. It is assumed that two sensors detect a common target, and only the offset detecting biases are considered.

Numerical examples

In this section, several numeral examples are used to evaluate the performance of the bias estimator with manoeuvring and non-manoeuvring targets.

Considering an example with two sensors and one target, the offset bias of sensor A is ξA(k)=500m,0.01rad,0.01radT. The stochastic noise has a zero mean and is white with a variance R(A)=diag1020.00220.0022. The sampling interval of sensor A is 0.09 s, and the offset bias of sensor B is ξB(k)=-750m,0.05rad,-0.02radT. The stochastic noise has a zero

Conclusions

In this paper, we proposed a novel solution to the bias estimation problem for asynchronous sensors. Using interpolation time registration, measurements from different sensors are modified to become coincident. Then, the pseudo measurement equation of the bias vector is obtained based on synchronous data. The primary advantage of the proposed algorithm is that the derivation of the pseudo measurement only depends on the target position; as a result, the proposed algorithm is not sensitive to

References (19)

  • Y.Y. Hu et al.

    Bias fusion estimation for multi-target tracking systems with multiple asynchronous sensors

    Aerospace Sci. Technol.

    (2013)
  • J.A. Besada et al.

    Generic multisensor multitarget bias estimation architecture

    IET Radar Sonar Navig.

    (2012)
  • Z.W. Zheng et al.

    New least squares registration algorithm for data fusion

    IEEE Trans. Aerospace Electron. Syst.

    (2004)
  • N. Nabaa et al.

    Solution to a multisensor tracking problem with sensor registration errors

    IEEE Trans. Aerospace Electron. Syst.

    (1999)
  • C.S. Hsieh

    General two-stage extended Kalman filters

    IEEE Trans. Autom. Control

    (2003)
  • B. Ristic et al.

    Sensor registration in ECEF coordinates using the MLR algorithm

  • J.A. Besada Portas et al.

    New approach to online optimal estimation of multisensor biases

    IEE Proc. Radar Sonar Navig.

    (2004)
  • X. Lin et al.

    Multisensor–multitarget bias estimation for general asynchronous sensors

    IEEE Trans. Aerospace Electron. Syst.

    (2005)
  • Y. Zhou et al.

    Sensor alignment with earth-centred earth-fixed (ECEF) coordinate systems

    IEEE Trans. Aerospace Electron. Syst.

    (1999)
There are more references available in the full text version of this article.

Cited by (8)

  • An overview of multirate multisensor systems: Modelling and estimation

    2019, Information Fusion
    Citation Excerpt :

    In addition, different from the above two modelling approaches, some additional modelling approaches are also investigated in recent years. Based on the idea of pseudo measurement approach, a B-spline interpolation time registration method is adopted to achieve synchronous data, and then an asynchronous bias estimation algorithm is presented by using the established pseudo measurement equation in [140]. Moreover, for systems with known exact fusion time, an asynchronous real-time track-to-track fusion algorithm is proposed in [141] based on the pseudosynchronised state estimates of local sensors, and so on.

  • State estimation for asynchronous sensor systems with Markov jumps and multiplicative noises

    2017, Information Sciences
    Citation Excerpt :

    For the multirate filter banks algorithm, a bank of multirate filters are designed and fused to achieve the global optimal estimates [14]. As for the distributed fusion approach, the optimal local estimators are designed at the state and measurement points by using the projection theory, and then the distributed optimal fusion estimator is given by using the matrix-weighted fusion rule in the LMMSE sense with the cross-covariance matrices between estimation errors of any two local estimators carefully computed [17,24,27,47]. In many practical systems, the parameter uncertainties can be modeled by multiplicative noises (also called state-dependent noises).

View all citing articles on Scopus
View full text