VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator

The monocular visual-inertial system (VINS), which consists one camera and one low-cost inertial measurement unit (IMU), is a popular approach to achieve accurate 6-DOF state estimation. However, such locally accurate visual-inertial odometry is prone to drift and cannot provide absolute pose estimation. Leveraging history information to relocalize and correct drift has become a hot topic. In this paper, we propose a monocular visual-inertial SLAM system, which can relocalize camera and get the absolute pose in a previous-built map. Then 4-DOF pose graph optimization is performed to correct drifts and achieve global consistent. The 4-DOF contains x, y, z, and yaw angle, which is the actual drifted direction in the visual-inertial system. Furthermore, the proposed system can reuse a map by saving and loading it in an efficient way. Current map and previous map can be merged together by the global pose graph optimization. We validate the accuracy of our system on public datasets and compare against other state-of-the-art algorithms. We also evaluate the map merging ability of our system in the large-scale outdoor environment. The source code of map reuse is integrated into our public code, VINS-Mono.

Automated driving requires highly precise and robust vehicle state estimation for its environmental perception, motion planning and control functions. Using GPS and environmental sensors can compensate for the deficits of the estimation based on traditional vehicle dynamics sensors. However, each type of sensor has specific strengths and limitations in accuracy and robustness due to their different properties regarding the quality of detection and robustness in diverse environmental conditions. For these reasons, we present a scalable concept for vehicle state estimation using an error-state extended Kalman filter (ESEKF) to fuse classical vehicle sensors with environmental sensors. The state variables, i.e., position, velocity and orientation, are predicted by a 6-degree-of-freedom (DoF) vehicle kinematic model that uses a low-cost inertial measurement unit (IMU) on a customer vehicle. The Error of the 6-DoF rigid body motion model is estimated using observations of global position using the global navigation satellite system (GNSS) and of the environment using radar, a camera and low-cost lidar. Our concept is scalable such that it is compatible with different sensor setups on different vehicle configurations. The experimental results compare various sensor combinations with measurement data in scenarios such as dynamic driving maneuvers on a test field. The results show that our approach ensures accuracy and robustness with redundant sensor data under regular and dynamic driving conditions.

A method of robust multi-rate state estimation

Delay-dependent robust asymptotic state estimation of Takagi–Sugeno fuzzy Hopfield neural networks with mixed interval time-varying delays

Nowadays, Autonomous Vehicles (AVs) are employed for various tasks, including spraying, harvesting, and planting.For AVs to navigate autonomously, accurate heading knowledge of the vehicle is essential.Sensors such as the Global Navigation Satellite System (GNSS) and Inertial Measurement Unit (IMU) are used on AVs to produce heading information.Dual-frequency and Real-Time Kinematic (RTK) capable systems with multiple antennas are used to increase the heading accuracy of GNSS.For IMUs, heading accuracy is directly related to the quality of the sensors, so high accuracy is achieved with expensive IMUs.However, with the development of micro-electro-mechanical system (MEMS) technology, studies are also being carried out on lowcost IMU solutions.This study tested the heading performances of three different sensors: a lowcost RTK/GNSS with multiple antennas, a tactical-grade IMU, and a low-cost MEMS IMU.An Unmanned Ground Agricultural Vehicle (UGAV) designed for spraying was driven on a line, and the sensors' data mounted on the UGAV were collected.Heading accuracy was also examined according to the distance between the RTK/GNSS system antennas.As a result of the analysis, the average errors of RTK/GNSS, tactical-grade IMU, and low-cost IMU are 0.58, 0.60, and 4.24 degrees, respectively.

Robust state estimation is the core capability for autonomous aerial robots operating in complex environments. Global navigation satellite system and visual odometry/SLAM are popular methods for state estimation. However, there exist scenarios, such as when flying between tall buildings or in the middle of deep canyons, that all these methods fail due to obstructed sky view and high altitude. in this letter, inspired by the fact that offline-generated three-dimensional (3-D) models of cities and natural scenes are readily available, we propose a global localization method for aerial robots by using 3-D models and measurements from a monocular fisheye camera and an inertial measurement unit (IMU). Due to the fact that 3-D models are generated by different cameras at different times, traditional feature-based or direct registration methods usually fail to perform, we therefore propose to use an edge alignment-based method for image-to-model registration under strong changes in lighting conditions and camera characteristics. We additionally aid our model-based localization with electronic image stabilization for better tracking performance, and extended Kalman filter (EKF)-based sensor fusion for closed-loop control. Our method runs onboard an embedded computer in real time. We verify both local accuracy and global consistency of the proposed approach in real-world experiments with comparisons against ground truth. We also show closed-loop control using the proposed method as state feedback.

Low-cost inertial measurement units (IMUs) based on microelectromechanical system (MEMS) have been widely used in self-localization for autonomous robots due to their small size and low power consumption. However, the low-cost MEMS IMUs often suffer from complex, non-linear, time-varying noise and errors. In order to improve the low-cost MEMS IMU gyroscope performance, a data-driven denoising method is proposed in this paper to reduce stochastic errors. Specifically, an attention-based learning architecture of convolutional neural network (CNN) and long short-term memory (LSTM) is employed to extract the local features and learn the temporal correlation from the MEMS IMU gyroscope raw signals. The attention mechanism is appropriately designed to distinguish the importance of the features at different times by automatically assigning different weights. Numerical real field, datasets and ablation experiments are performed to evaluate the effectiveness of the proposed algorithm. Compared to the raw gyroscope data, the experimental results demonstrate that the average errors of bias instability and angle random walk are reduced by 57.1 and 66.7%.

Inertial navigation systems for pedestrians are infrastructure-less and can achieve sub-meter accuracy in the short/medium period. However, when low-cost inertial measurement units (IMU) are employed for their implementation, they suffer from a slowly growing drift between the true pedestrian position and the corresponding estimated position. In this paper we illustrate a novel solution to mitigate such a drift by: a) using only accelerometer and gyroscope measurements (no magnetometers required); b) including the sensor error model parameters in the state vector of an extended Kalman filter; c) adopting a novel soft heuristic for foot stance detection and for zero-velocity updates. Experimental results evidence that our inertial-only navigation system can achieve similar or better performance with respect to pedestrian dead-reckoning systems presented in related studies, although the adopted IMU is less accurate than more expensive counterparts.

The use of microelectro-mechanical systems (MEMS)-based inertial measurement units (IMUs) is widespread in many applications concerning monitoring, diagnostic, and/or controlling in navigation and transportation systems, as well as in low-cost applications for automotive and aeronautical fields. The data provided by the set of sensors typically present in IMUs, as accelerometers, gyroscopes, and magnetometers, are often used also for feeding suitable filtering and positioning algorithms able to correct the attitude and path of the vehicle on which they are installed or to provide the analytical redundancy needed for online diagnosis. Nevertheless, on one hand, the performance of low-cost MEMS-based IMUs is certified only under a small set of nominal operating conditions, and on the other hand, the filtering algorithms are often designed and verified under canonical additive noises. In this framework, this article proposes a test plan and a test setup for analyzing and characterizing the performance of filtering algorithms for positioning based on data coming from low-cost IMUs and able to verify systematically the operation of such algorithms under real scenarios. Two kinds of very popular filtering algorithms have been considered, namely, the complementary filter and the attitude and heading reference systems (AHRS) Kalman filter, which belong to two opposite approaches. The experimental results prove how the typical vibrations present in real scenarios can significantly affect the performance of such algorithms.

State estimation in complex illumination environments based on conventional visual-inertial odometry is a challenging task due to the severe visual degradation of the visual camera. The thermal infrared camera is capable of all-day time and is less affected by illumination variation. However, most existing visual data association algorithms are incompatible because the thermal infrared data contains large noise and low contrast. Motivated by the phenomenon that thermal radiation varies most significantly at the edges of objects, the study proposes an ETIO, the first edge-based monocular thermal-inertial odometry, for robust state estimation in visually degraded environments. Instead of the raw image, we utilize the binarized image from edge extraction for pose estimation to overcome the poor thermal infrared image quality. Then, an adaptive feature tracking strategy ADT-KLT is developed for robust data association based on limited edge information and its distance distribution. Finally, a pose graph optimization performs real-time estimation over a sliding window of recent states by combining IMU pre-integration with reprojection error of all edge feature observations. We evaluated the performance of the proposed system on public datasets and real-world experiments and compared it against state-of-the-art methods. The proposed ETIO was verified with the ability to enable accurate and robust state estimation all-day time.

Sandia National Laboratories performs many expensive tests using inertial measurement units (IMUs)--systems that use accelerometers, gyroscopes, and other sensors to measure flight dynamics in three dimensions. For the purpose of this report, the metrics used to evaluate an IMU are cost, size, performance, resolution, upgradeability and testing. The cost of a precision IMU is very high and can cost hundreds of thousands of dollars. Thus the goals and results of this project are as follows: (1) Examine the data flow in an IMU and determine a generic IMU design. (2) Discuss a high cost IMU implementation and its theoretically achievable results. (3) Discuss design modifications that would save money for suited applications. (4) Design and implement a low cost IMU and discuss its theoretically achievable results. (5) Test the low cost IMU and compare theoretical results with empirical results. (6) Construct a more streamlined printed circuit board design reducing noise, increasing capabilities, and constructing a self-contained unit. Using these results, we can compare a high cost IMU versus a low cost IMU using the metrics from above. Further, we can examine and suggest situations where a low cost IMU could be used instead of a high cost IMU for saving cost, size, or both.

Delay-dependent robust exponential state estimation of Markovian jumping fuzzy Hopfield neural networks with mixed random time-varying delays

A robust distributed state estimation algorithm for AC/DC system with AC tie lines and high voltage direct current (LCC-HVDC) tie lines is proposed. The proposed algorithm consists of robust DC state estimation in the coordinator and robust distributed AC state estimation, and implements sequential solving by exchanging coupling variables of the AC/DC system. For AC state estimation, a distributed state estimation algorithm based on the bilinear algorithm is designed, which can achieve the same accuracy as the centralized algorithm. Robust state estimator based on the exponential absolute value function and exponential square function is suitable for reducing the influence of outliers and can reduce the operation scale by simplifying the nonlinear iterative process in distributed state estimation. Finally, the correctness and effectiveness of the algorithm is verified by simulation of three IEEE 118 bus systems interconnected through AC/DC hybrid tie lines.

There is a recognized complementarity between the vision system and the inertial measurement unit (IMU) in terms of autonomous navigation. Recently, visual inertial system (VINS) has become a hotspot of current research by fusing data of low-cost inertial measurement unit and vision system. However, the lack of direct measurement information makes the estimator’s initialization more difficult. This paper presents a method for online estimator initialization by using robust visual front end. One iterative process is employed to gradually align the vision system with inertial measurement unit. The convergence criterion can be used to determine the end of initialization, which can accurately recover parameters such as speed, scale, and gravity vector. This algorithm is applied to design a tightly coupled visual inertial odometry. In addition, experiments have been performed based on public data sets and equipment. The results show that the average absolute positioning error is less than 0.08 m, the relative positioning error is less than 0.03 m and the system has stable initialization performance with high accuracy positioning performance.KeywordsAutonomous navigationNonlinear optimizationVisual inertial system (VINS)Visual inertial odometer (VIO)

The smooth variable structure filter (SVSF) provides robust state estimation in case of model uncertainty but requires a linear measurement model and a full-rank measurement matrix. Doppler radar reports target position and range rate in the polar or spherical coordinates without the tangential velocity, leading to a nonlinear and underdetermined measurement model, which does not satisfy the requirements of the standard SVSF. In this paper, we propose a robust nonlinear estimator, named as the statically fused smooth variable structure filter (SFSVSF), for robust target state estimation with Doppler radar. First, the converted position measurement (CPM) is calculated in Cartesian coordinates and a converted Doppler measurement (CDM) is constructed from the product of the range and range rate measurements. Accordingly, the nonlinear Doppler radar measurement model is decomposed into a Cartesian-measurement model and a pseudo-measurement model. Both are linear time-invariant and underdetermined, and can be handled by the generalized SVSF. Thus, two generalized SVSFs, referred to as the CPM-SVSF and the CDM-SVSF, are implemented based on these two sub-models for robust estimations of the Cartesian state and the pseudo state, respectively. The Cartesian-state estimation and the pseudo-state estimation are statically fused by an approximate linear minimum mean square error (LMMSE) estimator for the final estimation of target state. Simulation results show the effectiveness of the SFSVSF algorithm in nonlinear estimation and the advantage in robustness over the EKF in case of target motion uncertainty.