For non-ellipsoidal extended object and group target tracking (NEOT NGTT), using a random matrix to simplify the extension as an ellipsoid, although efficient, may not be accurate enough because of loss useful information in shape orientation. In view this, we model or combination multiple ellipsoidal sub-objects, each represented by matrix. Different sub-objects are distinguished different dynamic models. Based on such models, Bayesian approach is proposed estimate kinematic states...

The approach of using a random matrix for extended object and group target tracking (EOT GTT) is appealing when the scattering centers or targets themselves are (partially) unresolvable. Designing effectively applying this relies on effective modeling group. To describe complex dynamical variation practical observation distortion extension in size, shape, orientation, two matrix-based models proposed. True measurement noise can also be incorporated into our proposed model easily. Facilitated...

The problem of modeling and estimation for linear equality constrained (LEC) systems is considered. exact dynamic model usually not readily available or too complicated, hence in many studies an auxiliary employed which the state does necessarily obey constraint strictly. Based on understanding that constraints, as prior information about state, should be incorporated into dynamics modeling, LEC (LECDM) constructed first. optimally fuses dynamics. Some its superior properties are presented....

For extended-object/group-target tracking (EOT/ GTT), the random-matrix approach assumes that measurements are linear in state and noise with a covariance being random matrix to represent object extension or target group. In practice, however, most nonlinear noise. This paper proposes for EOT/GTT using measurements. First, matched linearization (ML) is proposed linearize The linearized form has two parts. first state, it optimized sense of minimum mean square error (MMSE). second part...

For nonlinear estimation, the linear minimum mean square error (LMMSE) estimator using measurement augmented by its conversion can achieve better performance than original measurement. The main reason is that cannot be fully utilized LMMSE in a way. To effectively extract additional information which further estimator, approach named uncorrelated (UC) proposed. conversions of are with itself. Two specific approaches to generating UCs proposed based on Gaussian assumption and reference...

Large vision-language models struggle to accurately predict responses provided by multiple human annotators, particularly when those exhibit high uncertainty. In this study, we focus on a Visual Question Answering (VQA) task and comprehensively evaluate how well the output of state-of-the-art model correlates with distribution responses. To do so, categorize our samples based their levels (low, medium, high) uncertainty in disagreement (HUD) employ, not only accuracy, but also three new...

A multiple-model algorithm for maneuvering target tracking is proposed. It referred to as a second-order Markov chain (SOMC)-based interacting (SIMM) algorithm. The maneuver process modeled by SOMC incorporate more information. SIMM adopts merging strategy similar that of the (IMM) algorithm, except one-step model transition probabilities are updated based on SOMC. scheme proposed design tracking. performance evaluated via several scenarios Simulation results demonstrate effectiveness...

For extended object tracking (EOT), a fluctuating number of measurements are generated by sensor at time instant. In practice, the measurement depends on extension, resolution, and sensor-to-object geometry. Given number, thus, contains information state extension. This article proposes random-matrix approach to EOT utilizing this improve performance extension estimation. First, Gamma-alike distribution is proposed model dependence resolution also fits framework. Second, Bayesian jointly...

A new approach, referred to as best model augmentation (BMA), for variable-structure multiple-model (VSMM) estimation is presented. Here the original set of models augmented by a variable intended match unknown true mode. Based on Kullback-Leiber (KL) information, two versions criterion serving metric closeness between candidate and mode are derived in space states measurements, respectively. The adaptation (MSA) BMA turns out be an online optimization problem based KL criterion, which can...

The joint decision and estimation (JDE) algorithm is for solving problems involving simultaneous interdependent estimation. Based on the JDE approach with a generalized Bayes risk its recursive implementation (RJDE) dynamic system proposed recently, this paper proposes conditional (CJDE) risk, which generalization of risks conditioning data. We derive optimal solution that minimizes CJDE present an algorithm. For problems, version (RCJDE) also by following same spirit CJDE. To improve...

For extended object tracking, the random matrix approach is a computationally efficient framework that capable of estimating kinematic state, and extension jointly, thus gaining momentum in recent years. Existing approaches have an underlying assumption scatter centers are symmetrically distributed around centroid. In many real scenarios, however, they often on particular portions since these parts reflect more radar energy, measurement distributions over skewed. To effectively describe such...

Extended object tracking (EOT) is gaining momentum in recent years. The random matrix method a popular EOT method, which has simple yet effective framework. existing approaches usually assume that scatter centers are uniformly or symmetrically distributed on the object. However, due to weak signal strength and limited measuring angles, densely several (separated) parts rather than whole This results measurement-densely-distributed parts. To address this problem, paper proposes new...

Hidden Markov models are widely used for target tracking, where the process and measurement noises usually modeled as independent Gaussian distributions mathematical simplicity. However, independence assumptions do not always hold in practice. For example, a typical tracking application, radar is utilized to track non-cooperative target. Measurement noise correlated over time since sampling frequency of far greater than bandwidth noise. Besides, when maneuvering, heavy-tailed non-Gaussian...

For automotive radar-based extended object tracking, this paper proposes a new approach, which jointly estimates the kinematic state and extension of vehicle. The vehicle’s shape is described as rectangle with its vertices treated state. “Having rectangular shape” quadratic equality constraint on To deal challenging problem modeling measurement (scattering center) distribution over vehicle, we partition area into multiple regions assume that in each region scattering centers have simple...

A multiple conversion approach (MCA) to nonlinear estimation is proposed in this paper. It jointly considers hypotheses on the joint distribution of quantity be estimated and its measurement. The overall MCA estimate a probabilistically weighted sum hypothesis conditional estimates. To describe hypothesized distributions used match truth, general form characterized by (linear or nonlinear) measurement found. This more than Gaussian includes as special case. Moreover, minimum mean square...

Cheap, efficient, and industrial-grade model predictive controllers are required for extending the use of control (MPC) to resource-constrained embedded computing platforms practical industrial fields. In this paper, we implement matrix inversion on hardware improve computational efficiency our previously designed MPC controller. For specific matrices in each iteration active-set method, a simple positive definite symmetric algorithm is proposed, which can transform into iterated...

Direction of arrival (DOA) estimation underwater multipath signals has become an indispensable part many military and civil applications in the oceanic field. However, DOAs different paths may be same or too close to separated space, which brings a great challenge DOA signals. Besides, number is generally unknown time-varying passive scenarios lead underdetermined case. Therefore, it essential study efficient algorithm realize scenarios. This paper proposes via spatial time-frequency...

The multivariate autoregressive (MVAR) model is widely used in describing the dynamics of nonlinear systems, which estimates parameters and underlying states can be achieved by dual extended Kalman filter (DEKF). However, when measurements are corrupted complicated non-Gaussian noises, DEKF based on minimum mean-square error (MMSE) criterion may provide biased estimates. In present article, we develop a novel Kalman-type filter, referred to as under entropy (MEE) with fiducial points...

This paper considers tracking of extended objects using the measurements down-range and cross-range extent. type measurement can be naturally intuitively expressed in terms support functions. Based on functions, we propose a general approach to model smooth shapes objects. Another based Gaussian image is proposed nonsmooth such as polygons. Compared with existing approaches, larger range object modeled by which have concise mathematical forms favorable properties. Specifically for elliptical...

A variable-structure multiple-model (VSMM) approach, named equivalent-model augmentation (EqMA), is proposed. Here the model set augmented by a variable intended to best match unknown true mode. To fully utilize information provided sequences (model histories), this depends on mode at previous time. Thus different models correspond augmenting models. make estimation process computationally feasible, time approximated an equivalent (EqM) which provides closest results in sense of minimum...