Adaptive extended Kalman filtering for visual motion estimation of 3D objects

An algorithm for the real-time estimation of the position and orientation of a moving object of known geometry is presented in this paper. An estimation algorithm is adopted where a discrete-time extended Kalman filter computes the object pose on the basis of visual measurements of the object features. The scheme takes advantage of the prediction capability of the extended Kalman filter for the pre-selection of the features to be extracted from the image at each sample time. To enhance the robustness of the algorithm with respect to measurement noise and modelling error, an adaptive version of the extended Kalman filter, customized for visual applications, is proposed. Experimental results on a fixed single-camera visual system are presented to test the performance and the feasibility of the proposed approach.     

Comparison of stochastic filtering methods for 3D tracking

In the recent years, the 3D visual research has gained momentum with publications appearing for all aspects of 3D including visual tracking. This paper presents a review of the literature published for 3D visual tracking over the past five years. The work particularly focuses on stochastic filtering techniques such as particle filter and Kalman filter. These two filters are extensively used for tracking due to their ability to consider uncertainties in the estimation. The improvement in computational power of computers and increasing interest in robust tracking algorithms lead to increase in the use of stochastic filters in visual tracking in general and 3D visual tracking in particular. Stochastic filters are used for numerous applications in the literature such as robot navigation, computer games and behavior analysis. Kalman filter is a linear estimator which approximates system's dynamics with Gaussian model while particle filter approximates system's dynamics using weighted samples. In this paper, we investigate the implementation of Kalman and particle filters in the published work and we provide comparison between these techniques qualitatively as well as quantitatively. The quantitative analysis is in terms of computational time and accuracy. The quantitative analysis has been implemented using four parameters of the tracked object which are object position, velocity, size of bounding ellipse and orientation angle.     

Vehicle Localization in Outdoor Mountainous Forested Paths and Extension of Two-Dimensional Road Centerline Maps to Three-Dimensional Maps

This paper presents an approach for vehicle three-dimensional (3-D) localization in outdoor woodland environments where a previously available two-dimensional road centerline map is used in combination with a loosely coupled multi-sensor system to estimate the vehicle position in mountainous forested paths. The localization system is composed of a wheel encoder, an inertial measurement unit, a DGPS, a laser sensor and a barometer. An extended Kalman filter is used for sensor data fusion and pose estimation. When available, DGPS is used for 3-D dead reckoning accumulated error correction. During DGPS blackouts, the laser sensor is used for road extraction and measurement of the displacement of the vehicle to the road centerline, then the position is corrected towards the map. Moreover, the barometer that measures the height difference towards a reference is used to correct the estimated height in absence of DGPS 3-D data. The estimated height is added to the available road map to obtain a 3-D road centerline map that includes the road width measured with the laser sensor. Experimental results in large-scale real mountainous woodland environments show the robustness and simplicity of the proposed approach for vehicle localization and 3-D map extension.     

Localization of a wheeled mobile robot by sensor data fusion based on a fuzzy logic adapted Kalman filter

Accurate estimates of mobile robot location, if available, can be used to improve the performance of a vehicle dynamics control system. To this purpose, the data provided by odometric and sonar sensors are here fused together by means of an extended Kalman filter, providing robot position and orientation estimates at each sampling instant. To cope with the tracking of long trajectories, the performance of the filter is improved by introducing an on-line fuzzy-rule-based adaptation scheme.     

基于改进Kalman滤波的可见光极小目标TBD跟踪方法

针对无人机可见光图像极小目标跟踪问题,本文提出一种基于改进卡尔曼滤波的 (Tracking before detection,TBD)跟踪方法。首先利用检测算法定位目标位置作为卡尔曼滤波的测量值,检测过程中的匹配相似度参数作为卡尔曼滤波测量噪声协方差矩阵的参照依据,其次利用卡尔曼滤波建立跟踪框架预测下一帧的目标位置,最后检测模块以预测位置为 参考位置进行局部搜索,完成整个检测跟踪过程。为了提高跟踪效率,本文根据检测和预测位置积累误差判决检测模式,误差超过门限值则采取全局检测模式消除积累误差,否 则使用局部检测模式,降低TBD跟踪算法的运算复杂度。仿真实验证明,本文方法可以有效检测跟踪极小目标,提高跟踪的实时处理能力。     

Slip-compensated path following for planetary exploration rovers

A system that enables continuous slip compensation for a Mars rover has been designed, implemented and field-tested. This system is composed of several components that allow the rover to accurately and continuously follow a designated path, compensate for slippage and reach intended goals in high-slip environments. These components include visual odometry, vehicle kinematics, a Kalman filter pose estimator and a slip-compensated path follower. Visual odometry tracks distinctive scene features in stereo imagery to estimate rover motion between successively acquired stereo image pairs. The kinematics for a rocker–bogie suspension system estimates vehicle motion by measuring wheel rates, and rocker, bogie and steering angles. The Kalman filter processes measurements from an inertial measurement unit and visual odometry. The filter estimate is then compared to the kinematic estimate to determine whether slippage has occurred, taking into account estimate uncertainties. If slippage is detected, the slip vector is calculated by differencing the current Kalman filter estimate from the kinematic estimate. This slip vector is then used to determine the necessary wheel velocities and steering angles to compensate for slip and follow the desired path.     

Fusion of IMU and Vision for Absolute Scale Estimation in Monocular SLAM

The fusion of inertial and visual data is widely used to improve an object??s pose estimation. However, this type of fusion is rarely used to estimate further unknowns in the visual framework. In this paper we present and compare two different approaches to estimate the unknown scale parameter in a monocular SLAM framework. Directly linked to the scale is the estimation of the object??s absolute velocity and position in 3D. The first approach is a spline fitting task adapted from Jung and Taylor and the second is an extended Kalman filter. Both methods have been simulated offline on arbitrary camera paths to analyze their behavior and the quality of the resulting scale estimation. We then embedded an online multi rate extended Kalman filter in the Parallel Tracking and Mapping (PTAM) algorithm of Klein and Murray together with an inertial sensor. In this inertial/monocular SLAM framework, we show a real time, robust and fast converging scale estimation. Our approach does not depend on known patterns in the vision part nor a complex temporal synchronization between the visual and inertial sensor.     

Multi-sensor fusion for steerable four-wheeled industrial vehicles

This paper addresses the problem of multi-sensor data fusion in the navigation of a steerable four-wheeled industrial autonomous vehicle, which experiences substantial load variations of up to twice its weight. The practical considerations in the implementation of the filter are discussed. It aims to achieve a robust fusion algorithm with increased system tolerance against prolonged periods when absolute position updates are missing by improving estimation accuracy during dead-reckoning. The main contributions of this paper include the development of an adaptive estimator based on the extended Kalman filter to realise the multi-model filtering; the representation of the vehicle plant using a modified kinematic model to effectively describe the side-slip bias; the processing of redundant measurements to improve system immunity against noisy observations; and the ability to cope with periodically available odometry measurements and temporary position corrections from a landmark-based local reference system. To allow better adaptation to tyre wear and the wheels’ deflections under varying loads, the wheel encoder's resolution is constantly calibrated. The filter performance is evaluated at different speeds, loading patterns and maneuvers. Statistical tests are carried out to verify the filter consistency.     

紧耦合多传感器混合跟踪算法

在增强现实应用中实现对运动目标的准确跟踪是一个具有挑战性的任务。基于混合跟踪通过对多传感器信息的融合通常比单一传感器跟踪算法更为优越的特性,提出了一种新的紧耦合混合跟踪算法实现视觉与惯性传感器信息的实时融合。该算法基于多频率的测量数据同步,通过强跟踪滤波器引入时变衰减因子自适应调整滤波预测误差协方差,实现对运动目标位置数据的准确估计。通过标示物被遮挡状态下的跟踪实验结果表明,该方法能有效改善基于扩展卡尔曼滤波器的混合跟踪算法对运动目标位置信息预测估计的准确性,提高跟踪快速移动目标的稳定性,适用于大范围移动条件下的增强现实系统。     

A comparison of several nonlinear filters for reentry vehicle tracking

This paper compares the performance of several non-linear filters for the real-time estimation of the trajectory of a reentry vehicle from its radar observations. In particular, it examines the effect of using two different coordinate systems on the relative accuracy of an extended Kalman filter. Other filters considered are iterative-sequential filters, single-stage iteration filters, and second-order filters. It is shown that a range-direction-cosine extended Kalman filter that uses the measurement coordinate system has less bias and less rms error than a Cartesian extended Kalman filter that uses the Cartesian coordinate system. This is due to the fact that the observations are linear in the range-direction-cosine coordinate system, but nonlinear in the Cartesian coordinate system. It is further shown that the performance of the Cartesian iterative-sequential filter that successively relinearizes the observations around their latest estimates and that of a range-direction-cosine extended Kalman filter are equivalent to first order. The use of a single-stage iteration to reduce the dynamic nonlinearity improves the accuracy of all the filters, but the improvement is very small, indicating that the dynamic nonlinearity is less significant than the measurement nonlinearity in reentry vehicle tracking under the assumed data rates and measurement accuracies. The comparison amongst the nonlinear filters is carried out using ten sets of observations on two typical trajectories. The performance of the filters is judged by their capability to eliminate the initial bias in the position and velocity estimates.     

Emitter location using an angle-only tracking filter with own-ship trajectory estimation

An angle-only tracking algorithm to locate the emitter source position from an overflight vehicle is presented. The algorithm uses an extended Kalman filter with angular data from an onboard direction finder. The dynamic relationship between the emitter and own-ship motion is formulated in modified polar coordinates (MPC), which yields good noise-handling performance. The MPC filter method, however, encounters slow convergence problem under realistic overflight scenarios, where the lateral sightline motion inputs are mild. By using the knowledge on an own-ship estimator and computed pseudo-measurements for range and range-rate over range, the convergence of estimator is greatly accelerated. The combined scheme of this trajectory estimation filter and the MPC filter markedly improves the tracking accuracy as well. Simulation results reveal that the proposed algorithm is superior to that of the MPC filter algorithm.     

Mapping extended Kalman filters onto linear arrays

Techniques for mapping extended Kalman filters onto linear arrays of programmable cells designed for real-time applications are described. First, a general method for mapping a standard (nonsquare root) Kalman filter, where the columns of the covariance matrix are updated in parallel, is introduced. Next, a general method for mapping a factorized (square root) filter, where fast Givens rotations are used to triangularize the prematrix and where rotations of the rows of the prematrix are performed in parallel, is introduced. These mappings are used to implement an extended Kalman filter commonly used in target tracking applications on the Warp computer. The Warp is a commercially available linear array of 10 or more programmable cells connected to an MC68020-based workstation. The Warp implementation of the standard Kalman filter running on 8 Warp cells achieves a measured speedup of 7 over the same filter running on a single cell. The Warp implementation of the factorized filter running on 10 Warp cells achieves a measured speedup of 2     

The Motor Extended Kalman Filter: A Geometric Approach for Rigid Motion Estimation

In this paper the motor algebra for linearizing the 3D Euclidean motion of lines is used as the oretical basis for the development of a novel extended Kalman filter called the motor extended Kalman filter (MEKF). Due to its nature the MEKF can be used as online approach as opposed to batch SVD methods. The MEKF does not encounter singularities when computing the Kalman gain and it can estimate simultaneously the translation and rotation transformations. Many algorithms in the literature compute the translation and rotation transformations separately. The experimental part demonstrates that the motor extended Kalman filter is an useful approach for estimation of dynamic motion problems. We compare the MEKF with an analytical method using simulated data. We present also an application using real images of a visual guided robot manipulator; the aim of this experiment is to demonstrate how we can use the online MEKF algorithm. After the system has been calibrated, the MEKF estimates accurately the relative position of the end-effector and a 3D reference line. We believe that future vision systems being reliably calibrated will certainly make great use of the MEKF algorithm.     

基于多传感器的智能车辆姿态解算方法

针对智能车辆主动环境感知的需求,提出了一种采用三轴加速度计、三轴磁强计和三轴陀螺仪组合进行车辆姿态解算的方法.首先以旋转矢量法为陀螺仪的车辆姿态解算方法,作为扩展卡尔曼滤波的状态方程,用于车辆姿态的预测;其次以高斯牛顿法为加速度计和磁强计的车辆姿态解算方法,作为扩展卡尔曼滤波的观测方程,用于车辆姿态校正;然后在此基础上构建扩展卡尔曼滤波传播方程,采用扩展卡尔曼滤波进行多传感器信息融合,得到车辆的姿态解算结果;最后通过构建实车测试环境对解算方法有效性进行验证.实验结果表明,通过基于多传感器的车辆姿态解算方法解算得到的车辆姿态角稳定、准确,能够满足智能车辆行为参数估计的实际需求.     

Adaptive intelligent cascade control of a ball-riding robot for optimal balancing and station-keeping

This paper presents an adaptive intelligent cascade control strategy to maintain the dynamic stability of a ball-riding robot (BRR). The four-wheeled mechanism beneath the robot body balances it on a spherical wheel. The BRR is modeled as a combination of two decoupled inverted pendulums. Therefore, two independent controllers are used to control its pitch and roll rotations. An incremental proportional–integral–derivative (PID) is implemented in the inner loop of the cascade to maintain the vertical balance. A generic PD controller is used in the outer loop to keep the station by controlling its spatial position. The controller parameters are automatically tuned via a fuzzy adaptation mechanism. The centers of fuzzy output membership functions are dynamically updated via an extended Kalman filter (EKF). The proposed controller quickly responds to changes in system’s state and effectively rejects the exogenous disturbances. The results of real-time experiments are presented to validate the effectiveness of the proposed hybrid controller over the conventional classical controllers.     

Tracking of multiple target types with a single neural extended Kalman filter

The neural extended Kalman filter is an adaptive state estimation routine that can be used in target‐tracking systems to aid in the tracking through maneuvers without prior knowledge of the targets' dynamics. Within the neural extended Kalman filter, a neural network is trained using a Kalman filter training paradigm that is driven by the same residual as the state estimator. The difference between the a priori model used in the prediction steps of the estimator and the actual target dynamics is approximated. An important benefit of the technique is its versatility because little if any a priori knowledge of the target dynamics is needed. This allows the technique to be used in a generic tracking system that will encounter various classes of targets. In this paper, the neural extended Kalman filter is applied simultaneously to three separate classes of targets, each with different maneuver capabilities. The results show that the approach is well suited for use within a tracking system with multiple possible or unknown target characteristics. © 2010 Wiley Periodicals, Inc.     

A Dynamic-Model-Based Wheel Slip Detector for Mobile Robots on Outdoor Terrain

This paper introduces a model-based approach to estimating longitudinal wheel slip and detecting immobilized conditions of autonomous mobile robots operating on outdoor terrain. A novel tire traction/braking model is presented and used to calculate vehicle dynamic forces in an extended Kalman filter framework. Estimates of external forces and robot velocity are derived using measurements from wheel encoders, inertial measurement unit, and GPS. Weak constraints are used to constrain the evolution of the resistive force estimate based upon physical reasoning. Experimental results show the technique accurately and rapidly detects robot immobilization conditions while providing estimates of the robot's velocity during normal driving. Immobilization detection is shown to be robust to uncertainty in tire model parameters. Accurate immobilization detection is demonstrated in the absence of GPS, indicating the algorithm is applicable for both terrestrial applications and space robotics.      

改进强跟踪滤波算法及其在汽车状态估计中的应用

准确实时地获取汽车行驶过程中的状态变量,对汽车底盘控制有着重要的意义,而这些关键状态往往难以直接测量或 者成本较高.结合纵向、侧向和横摆三自由度非线性汽车模型,将改进强跟踪滤波(Improved strong track filter, ISTF)算法应用到汽车的状态估计中,并改进了算 法的稳定性.与扩展卡尔曼滤波(Extended Kalman filter, EKF)算法进行比较分析.通过Carsim和Matlab/Simulink联合仿真和实车双移线实验验证算法,结果 表明,该算法在估计精度、跟踪速度、抑制噪声等方面均优于扩展卡尔曼滤波算法,满足汽车状态估计器的软件性能要求.     

A robust optical/inertial data fusion system for motion tracking of the robot manipulator

We present an optical/inertial data fusion system for motion tracking of the robot manipulator, which is proved to be more robust and accurate than a normal optical tracking system（OTS）. By data fusion with an inertial measurement unit（IMU）, both robustness and accuracy of OTS are improved. The Kalman filter is used in data fusion. The error distribution of OTS pro-vides an important reference on the estimation of measurement noise using the Kalman filter. With a proper setup of the system and an effective method of coordinate frame synchronization, the results of experiments show a significant improvement in terms of robustness and position accuracy.