Unscented Particle Filter Research Articles

Improved particle filter based on differential evolution

Resampling schemes for a particle filter based on the differential evolution (DE) algorithm are presented. By using these schemes, several types of differential evolution particle filters (DEPFs) are proposed. In the proposed filters, the unscented Kalman filter is utilised to generate the importance proposal distribution and the different DE algorithms are used as the resampling scheme. Simulation results demonstrate that the proposed DEPFs outperform the sequential importance resampling algorithm, the regularised particle filter, and the unscented particle filter.

Practical issues in the application of the particle filter for estimation of chemical processes

AbstractIn this work, we explore practical aspects of the application of particle filters to chemical processes, with special emphasis on polymerization processes. Particle filters are potentially better suited for estimation of these nonlinear processes than Kalman filter-based estimators, since they do not make the assumption that the process and measurement noise are Gaussian. We demonstrate the improved performance of particle filters over the extended Kalman filter and the unscented Kalman filter when the innovation sequences are non-Gaussian. Another potential advantage of the particle filter is that information on full distribution of the state is obtained, and not just the expectation of the state estimate. We provide results based on k-means clustering that indicate the best method of extracting a point estimate from the full state distribution. Finally, we show that in cases where there is high plant-model mismatch, e.g., when a reduced order model is used for estimation, the basic sequential importance resampling particle filter (SIR-PF) does not provide accurate estimates, while an unscented particle filter provides excellent estimates even under high structural plant-model mismatch.

Application Of Extended Kalman Filter For A Free Falling Body Towards Earth

State estimation theory is one of the best mathematical approaches to analyze variants in the states of the system or process. The state of the system is defined by a set of variables that provide a complete representation of the internal condition at any given instant of time. Filtering of Random processes is referred to as Estimation, and is a well-defined statistical technique. There are two types of state estimation processes, Linear and Nonlinear. Linear estimation of a system can easily be analyzed by using Kalman Filter (KF) and is used to compute the target state parameters with a priori information under noisy environment. But the traditional KF is optimal only when the model is linear and its performance is well defined under the assumptions that the system model and noise statistics are well known. Most of the state estimation problems are nonlinear, thereby limiting the practical applications of the KF. The modified KF, aka EKF, Unscented Kalman filter and Particle filter are best known for nonlinear estimates. Extended Kalman filter (EKF) is the nonlinear version of the Kalman filter which linearizes about the current mean and covariance. The estimation can be linearised around the current estimate using the partial derivatives to compute estimates even in the face of nonlinear relationships.. The EKF has been considered the standard in the theory of nonlinear state estimation. This paper deals with how to estimate a nonlinear model with Extended Kalman filter (EKF). The approach in this paper is to analyze Extended Kalman filter where EKF provides better probability of state estimation for a free falling body towards earth.

Rao-Blackwellised unscented particle filtering for jump Markov non-linear systems: an H∞ approach

In this study, a robust Rao-Blackwellised particle filter (RBPF) is proposed for jump Markov non-linear systems (JMNLSs) with unknown noise statistics. A non-linear filter is presented by applying the unscented transform technique in the H∞ setting, which is used to update the continuous-state particles in the RBPF framework. Moreover, a way to adaptively adjust the disturbance tolerance level for performance requirement is presented. Simulation results using the proposed approach are also presented.

Online identification of nonlinear multivariable processes using self-generating RBF neural networks

This paper addresses the problem of online model identification for multivariable processes with nonlinear and time-varying dynamic characteristics. For this purpose, two online multivariable identification approaches with self-organizing neural network model structures will be presented. The two adaptive radial basis function (RBF) neural networks are called as the growing and pruning radial basis function (GAP-RBF) and minimal resource allocation network (MRAN). The resulting identification algorithms start without a predefined model structure and the dynamic model is generated autonomously using the sequential input-output data pairs in real-time applications. The extended Kalman filter (EKF) learning algorithm has been extended for both of the adaptive RBF-based neural network approaches to estimate the free parameters of the identified multivariable model. The unscented Kalman filter (UKF) has been proposed as an alternative learning algorithm to enhance the accuracy and robustness of nonlinear multivariable processes in both the GAP-RBF and MRAN based approaches. In addition, this paper intends to study comparatively the general applicability of the particle filter (PF)-based approaches for the case of non-Gaussian noisy environments. For this purpose, the Unscented Particle Filter (UPF) is employed to be used as alternative to the EKF and UKF for online parameter estimation of self-generating RBF neural networks. The performance of the proposed online identification approaches is evaluated on a highly nonlinear time-varying multivariable non-isothermal continuous stirred tank reactor (CSTR) benchmark problem. Simulation results demonstrate the good performances of all identification approaches, especially the GAP-RBF approach incorporated with the UKF and UPF learning algorithms. Copyright © 2010 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society

Speed Estimation from Single Loop Data Using an Unscented Particle Filter

Abstract: This article presents a hybrid method, the Unscented Particle Filter (UPF), for traffic flow speed estimation using single loop outputs. The Kalman filters used in past speed estimation studies employ a Gaussian assumption that is hardly satisfied. The hybrid method that combines a parametric filter (Unscented Kalman Filter) and a nonparametric filter (Particle Filter) is thus proposed to overcome the limitations of the existing methods. To illustrate the advantage of the proposed approach, two data sets collected from field detectors along with a simulated data set are utilized for performance evaluation and comparison with the Extended Kalman Filter and the Unscented Kalman Filter. It is found that the proposed method outperforms the evaluated Kalman filter methods. The UPF method produces accurate speed estimation even for congested flow conditions in which many other methods have significant accuracy problems.

Mobile Robot Monte Carlo Localization Based on Improved Unscented Particle Filter

摘要: 粒子滤波是移动机器人蒙特卡罗定位(Monte Carlo localization, MCL)的核心环节. 首先, 针对粒子滤波过程的粒子退化问题, 利用迭代Sigma点卡尔曼滤波来精确设计粒子滤波器的提议分布, 以迭代更新方式将当前观测信息融入顺序重要性采样过程, 提出IUPF (Improved unscented particle filter)算法. 然后, 将IUPF与移动机器人MCL相结合, 给出IUPF-MCL定位算法的实现细节. 仿真结果表明, IUPF-MCL是一种精确鲁棒的移动机器人定位算法. 关键词: 移动机器人 / 蒙特卡罗定位 / 粒子滤波 / 无迹卡尔曼滤波

Federated unscented particle filtering algorithm for SINS/CNS/GPS system

To solve the problem of information fusion in the strapdown inertial navigation system (SINS)/celestial navigation system (CNS)/global positioning system (GPS) integrated navigation system described by the nonlinear/non-Gaussian error models, a new algorithm called the federated unscented particle filtering (FUPF) algorithm was introduced. In this algorithm, the unscented particle filter (UPF) served as the local filter, the federated filter was used to fuse outputs of all local filters, and the global filter result was obtained. Because the algorithm was not confined to the assumption of Gaussian noise, it was of great significance to integrated navigation systems described by the non-Gaussian noise. The proposed algorithm was tested in a vehicle’s maneuvering trajectory, which included six flight phases: climbing, level flight, left turning, level flight, right turning and level flight. Simulation results are presented to demonstrate the improved performance of the FUPF over conventional federated unscented Kalman filter (FUKF). For instance, the mean of position-error decreases from (0.640×10−6 rad, 0.667×10−6 rad, 4.25 m) of FUKF to (0.403×10−6 rad, 0.251×10−6 rad, 1.36 m) of FUPF. In comparison of the FUKF, the FUPF performs more accurate in the SINS/CNS/GPS system described by the nonlinear/non-Gaussian error models.

TRACKING ALGORITHM FOR MMW/IR MULTI-SENSORS FUSION

Millimeter wave(MMW)/infrared (IR) sensor is a key technology for composite guidance system of missiles. Aimming to solve the problems that there were linear errors in the algorithm of square-root unscented Kalman filter (SR-UKF) and it was difficult to obtain the importance density function for the algorithm of particle filter(PF), a square-root unscented Kalman particle filter (SR-UK-PF) algorithm with the sequential fusion was presented by combining SR-UKF with PF. The main idea of this algorithm was to calculate the state transition matrix and the error covariance matrix by SR-UKF, and to construct the importance density function by the sequential fusion of particle filter. Thus, the importance density function could integrate the latest observation into system state transition density, and the proposal distribution could be more in line with the distribution of real states. To demonstrate the effectiveness of this model, simulations were carried out based on tracking algorithm for the surface-to-air missile with MMW/IR sensor.The results show that this technique can overcome the flaw that it is hard to get the optimization importance density function in the particle filter, and it can significantly improve the accuracy of state estimation for the system with multi-sensors.

An Information-Theoretic Approach for Energy-Efficient Collaborative Tracking in Wireless Sensor Networks

The problem of collaborative tracking of mobile nodes in wireless sensor networks is addressed. By using a novel metric derived from the energy model in LEACH (W.B. Heinzelman, A.P. Chandrakasan and H. Balakrishnan, Energy-Efficient Communication Protocol for Wireless Microsensor Networks, in: Proceedings of the 33rd Hawaii International Conference on System Sciences (HICSS '00), 2000) and aiming at an efficient resource solution, the approach adopts a strategy of combining target tracking with node selection procedures in order to select informative sensors to minimize the energy consumption of the tracking task. We layout a cluster-based architecture to address the limitations in computational power, battery capacity and communication capacities of the sensor devices. The computation of the posterior Cramer-Rao bound (PCRB) based on received signal strength measurements has been considered. To track mobile nodes two particle filters are used: the bootstrap particle filter and the unscented particle filter, both in the centralized and in the distributed manner. Their performances are compared with the theoretical lower bound PCRB. To save energy, a node selection procedure based on greedy algorithms is proposed. The node selection problem is formulated as a cross-layer optimization problem and it is solved using greedy algorithms.

An improved particle filtering algorithm based on observation inversion optimal sampling

According to the effective sampling of particles and the particles impoverishment caused by re-sampling in particle filter, an improved particle filtering algorithm based on observation inversion optimal sampling was proposed. Firstly, virtual observations were generated from the latest observation, and two sampling strategies were presented. Then, the previous time particles were sampled by utilizing the function inversion relationship between observation and system state. Finally, the current time particles were generated on the basis of the previous time particles and the system one-step state transition model. By the above method, sampling particles can make full use of the latest observation information and the priori modeling information, so that they further approximate the true state. The theoretical analysis and experimental results show that the new algorithm filtering accuracy and real-time outperform obviously the standard particle filter, the extended Kalman particle filter and the unscented particle filter.

Small target track-before-detect algorithm based on unscented particle filtering

In order to improve the joint detecting and tracking ability for small targets under complex environments,the authors presented a new Track-Before-Detect (TBD) algorithm based on particle filter. Unscented Particle Filter (UPF) method was adopted and the motion status and appearance status were jointly scanned in the particle filter. The new algorithm essentially introduces tracking idea into target detecting. It has the merits of high-sensitivity,high acquisition probability,good tracking performance and low requirement on sample particle number. The algorithm can effectively enhance the tracking-while-detecting ability of infrared searching and tracking system.

Reliable people tracking approach for mobile robot in indoor environments

There is a great challenge that a mobile robot reliably and continuously tracks a specific person in indoor environments. In this paper, a novel method is presented, which can effectively recognize and reliably track a target person based on mobile robot vision. Gabor wavelet and hidden Markov model (HMM) are firstly employed for identifying the target person. In order to effectively track the specific person and reduce the computational cost in tracking stage, horizontal-projecting probability histogram (HPPH) of upper body color clothes region is proposed for extracting the pattern features, which significantly improves the tracking reliability and, at the same time, unscented particle filter (UPF) is integrated and PID operator is introduced for controlling the robot to follow the person. Experimental results validate the robustness and the reliability of this approach.

퍼지추론/UPF를 이용한 UGV의 GPS/INS 데이터 융합 및 위치추정

A GPS/INS system is widely used in the UGV to estimate position during the mission. However, there are few restrictions when a GPS/INS system used alone. For example, GPS provides precise location information but easily interrupted by external factors like weather, environment, etc. INS provides continuous location data but positioning errors grew very rapidly with time. Therefore, it is necessary to integrating multi-sensors for more continuous and correct position estimation. In this paper, we propose location estimation algorithm of the UGV for GPS/INS integrated system that combines Fuzzy Inference and Unscented Particle Filter(UPF) to improve navigation. Fuzzy inference provides the simplest method integrating GPS/INS and UPF is non-linear estimation filter well suited to the correction of errors. The performance of the proposed algorithm was tested to compare with other algorithms. the results show that the proposed algorithm is more accuracy in position estimation and ensures continuous position tracking.

Improved Filter Strategies for Precise Geolocation of Unexploded Ordnance using IMU/GPS Integration

Efficient and precise geolocation can be achieved by integrating a ranging system, such as GPS, with inertial sensors in order to bridge short outages, enhance accuracy degradation, and increase the temporal resolution in the ranging system. Optimal integration depends on appropriate filter methods that can accommodate the particular short-term dynamics experienced by platforms, such as UXO ground-based detection systems. The traditional extended Kalman filter was designed to integrate data from a linearized system excited by Gaussian noise. We compared this filter to modern filters that obviate these prerequisites, including the unscented Kalman filter, the particle filter, and adaptive variations thereof, using simulated IMU/ranging systems that follow a typical trajectory with both straight and curved segments. The unscented filter performed significantly better than the extended Kalman filter, particularly over the curved segments, yielding up to 50% improvement in the position accuracy using medium-grade inertial measurement units. Similar improvement was obtained for the unscented particle filter, and its adaptive variant, over the unscented Kalman filter (which performed comparably to the extended Kalman filter) when the statistical distribution of the IMU noise was non-symmetric (i.e., essentially non-Gaussian). While the few-centimetre geolocation accuracy goal for highly dynamic UXO characterization applications remains a challenge if tactical grade IMUs are integrated with a significantly degraded ranging system, using filters appropriate to the inherent nonlinear dynamics and potential non-Gaussian nature of the sensor noise tend to reduce overall errors compared to the traditional filter.

An Ultrasound-based Localization Algorithm for Catheter Ablation Guidance in the Left Atrium

We present a method for catheter localization in the left atrium based on the unscented particle filter (UPF), a Monte Carlo method employed in stochastic state estimation. Using an intracardiac echo (ICE) ultrasound catheter, we acquire ultrasound images of the atrium from multiple configurations and iteratively determine the catheter’s pose with respect to anatomy. At each time step, the catheter’s change in pose is determined using either a six-degree-of-freedom electromagnetic pose sensor or a robotic guide catheter whose kinematics serve as a pseudo-pose measurement. Sensor and kinematic model uncertainties are explicitly considered when computing the localization estimate. Acquired ultrasound images are compared with simulated ultrasound images based on segmented computed tomography (CT) or magnetic resonance (MR) data of the left atrium. The results of these comparisons are used to refine the localization estimate. After considering less than 30 seconds’ worth of ICE data, our algorithm converges to an accurate pose estimate. Furthermore, our algorithm is robust to sensor drift and kinematic model errors, as well as gradual, unmodeled movements in the anatomy. Such problems typically complicate traditional image-based localization.

Variance reduction techniques in particle-based visual contour tracking

This paper presents a comparative study of three different strategies to improve the performance of particle filters, in the context of visual contour tracking: the unscented particle filter, the Rao-Blackwellized particle filter, and the partitioned sampling technique. The tracking problem analyzed is the joint estimation of the global and local transformation of the outline of a given target, represented following the active shape model approach. The main contributions of the paper are the novel adaptations of the considered techniques on this generic problem, and the quantitative assessment of their performance in extensive experimental work done.

Realization of an autonomous integrated suite of strapdown astro-inertial navigation systems using unscented particle filtering

Multisensor navigation data synthesis (MNDS) is the process of fusing outputs from inertial sensors with information from other sensors and information processing blocks into one representational form. This technique is anticipated to accomplish enhanced accuracy and more specific inferences than could be achieved by the use of a single sensor alone. Therefore, this research work expounds innovative filtering methodology for the multisensor navigation data synthesis for a ballistic missile application that augments navigation system performance. The premise and characteristics of strapdown inertial navigation system (SINS) integrated with the astronavigation system (ANS) based on the unscented particle filter (UPF) are investigated in this paper. Configuration of the integrated navigation system is presented with its canonical model, and system dynamic and stochastic models required for the filtering algorithm are presented. To exemplify integrated navigation filter mechanization is the foremost aspiration of this research. To validate and corroborate the designed MNDS technique, simulations are carried out that demonstrate the validity of this method on enhancing the navigation system’s accuracy with estimation and compensation for the gyro’s drift. This integrated system results in a significant reduction in impact-point dispersion of a re-entry vehicle.

Spacecraft autonomous navigation using unscented particle filter-based celestial/Doppler information fusion

With the development of space missions, especially manned space missions, a reliable and secure navigation system, and assured autonomous capability in case of emergencies in space, is needed. In order to compensate for the insufficiency of ground station tracking and control, a new autonomous celestial/Doppler-integrated navigation method for a spacecraft is proposed. Celestial navigation is a fully autonomous navigation method, but in some situations the navigation accuracy of this method is subject to the inaccuracies of the measuring devices. Doppler navigation can serve as a good complement to celestial navigation. Because both the state and the measurement models of a celestial/Doppler- integrated navigation system are nonlinear and non-Gaussian, the unscented particle filter (UPF) based information fusion method is proposed here to fuse the position signals from the celestial navigation and Doppler navigation subsystems, and to enhance the navigation accuracy. The performance of this new method is tested and examined using actual spacecraft-orbit data. Simulations show that the position and velocity accuracies are estimated to within 300 m and 0.5 m s−1 respectively, which demonstrate the feasibility and effectiveness of this method. Moreover, it can be used as a backup system to provide redundancy.

UPF based autonomous navigation scheme for deep space probe

The autonomous “celestial navigation scheme” for deep space probe departing from the earth and the autonomous “optical navigation scheme” for encountering object celestial body are presented. Then, aiming at the conditions that large initial estimation errors and non-Gaussian distribution of state or measurement errors may exist in orbit determination process of the two phases, UPF (unscented particle filter) is introduced into the navigation schemes. By tackling nonlinear and non-Gaussian problems, UPF overcomes the accuracy inuence brought by the traditional EKF (extended Kalman filter), UKF (unscented Kalman filter), and PF (particle filter) schemes in approximate treatment to nonlinear and non-Gaussian state model and measurement model. The numerical simulations demonstrate the feasibility and higher accuracy of the UPF navigation scheme.