In the near future, multi-sensor fusion will be the core component to navigate the autonomous driving platforms. This paper proposes new strategies to cope with the integration of an inertial navigation system (INS), a global navigation satellite system (GNSS), and light detection and ranging (LIDAR) to achieve simultaneous localization and mapping (INS/GNSS/LiDAR SLAM) especially in GNSS challenging environments where GNSS signals are blocked or contaminated with reflected signals. The proposed strategies implement a high level of integration with various information received from multiple sensors to collectively compensate for the specific drawbacks of those sensors included in the integrated system. The first strategy is to solve the divergence and drift problems of SLAM using the initial pose information from INS and the proposed refreshing process using an INS/GNSS integrated system. In addition, an updated mechanization is designed to qualify those received measurements based on cross validation of separate types of data. This mechanization is to ensure all measurements are reliable for the Extended Kalman Filter (EKF) update process. Moreover, the SLAM-derived information plays a major role to recognize the vehicle movement which assists the system to accurately apply those appropriate vehicle motion constraint models. The preliminary results presented in this study illustrate that proposed algorithm performs superior than the traditional INS/GNSS integration scheme and provides absolute navigation accuracy of 2 meters and 0.6% of distance traveled in GNSS-denied as well as 1.2 meters in GNSS-hostile environments, respectively.
Global Navigation Satellite Systems (GNSS) integrated with multiple sensors have been widely applied in many Intelligent Transport Systems (ITS). Intelligent vehicles, increasingly a key component of the future transport system, require high-performance positioning, navigation and timing (PNT) technologies. This cannot be achieved by current 2-dimnsional systems. This is because accurate comprehensive state information (time, position, derivatives and attitude) is required to estimate amongst others, the vehicle’s power usage as well as to enable precise environmental mapping based on accurate state of the relevant sensors and path planning including in complex multi-level intersections. Because of the multipath effects and signal interruption in urban environments, comprehensive vehicle state estimation is not always available at the required level of performance. To address this issue, we propose an effective method to integrate the Inertial Measurement Unit (IMU), Global Positioning System (GPS) and monocular Visual Odometry (VO) for urban vehicle navigation. A robust Extended Kalman Filter (EKF) based two-step integration algorithm is developed with a non-holonomic constraint (NHC). In particular, the NHC is not only applied on the offline VO error modelling process, but also on the online sensor fusion process to improve the 3D vehicle state estimation. The proposed IMU/GPS/VO integration scheme is tested with various sensor levels in different urban environments. The results show that the proposed IMU/GPS/VO fusion algorithm could deliver a 3D RMSE of 3.285m, which outperforms the other conventional candidate fusion schemes in the noisy GNSS urban areas. The further test in the urban with outages has demonstrated that the proposed algorithm delivers an overall 3D RMSE of 1.290m, 0.073m/s and 0.486 degrees, in terms of positioning, velocity and attitude, respectively. It is also demonstrated that the proposed low-cost VO integration with IMU/GPS could achieve similar performance with the high-cost odometer based integration in deep urban areas but with advantages of higher flexibility and lower cost.
Automated driving has made considerable progress recently. The multisensor fusion system is a game changer in making self-driving cars possible. In the near future, multisensor fusion will be necessary to meet the high accuracy needs of automated driving systems. This paper proposes a multisensor fusion design, including an inertial navigation system (INS), a global navigation satellite system (GNSS), and light detection and ranging (LiDAR), to implement 3D simultaneous localization and mapping (INS/GNSS/3D LiDAR-SLAM). The proposed fusion structure enhances the conventional INS/GNSS/odometer by compensating for individual drawbacks such as INS-drift and error-contaminated GNSS. First, a highly integrated INS-aiding LiDAR-SLAM is presented to improve the performance and increase the robustness to adjust to varied environments using the reliable initial values from the INS. Second, the proposed fault detection exclusion (FDE) contributes SLAM to eliminate the failure solutions such as local solution or the divergence of algorithm. Third, the SLAM position velocity acceleration (PVA) model is used to deal with the high dynamic movement. Finally, an integrity assessment benefits the central fusion filter to avoid failure measurements into the update process based on the information from INS-aiding SLAM, which increases the reliability and accuracy. Consequently, our proposed multisensor design can deal with various situations such as long-term GNSS outage, deep urban areas, and highways. The results show that the proposed method can achieve an accuracy of under 1 meter in challenging scenarios, which has the potential to contribute the autonomous system.
21 August 2017 in IEEE Geoscience and Remote Sensing Letters
In disaster management, reconstructing the environment and quickly collecting the geospatial data of the impacted areas in a short time are crucial. In this letter, a light detection and ranging (LiDAR)-based unmanned aerial vehicle (UAV) is proposed to complete the reconstruction task. The UAV integrate an inertial navigation system (INS), a global navigation satellite system (GNSS) receiver, and a low-cost LiDAR. An unmanned helicopter is introduced and the multisensor payload architecture for direct georeferencing is designed to improve the capabilities of the vehicle. In addition, a new strategy of iterative closest point algorithm is proposed to solve the registration problems in the sparse and inhomogeneous derived point cloud. The proposed registration algorithm addresses the local minima problem by the use of direct-georeferenced points and the novel hierarchical structure as well as taking the feedback bias into INS/GNSS. The generated point cloud is compared with a more accurate one derived from a high-grade terrestrial LiDAR which uses real flight data. Results indicate that the proposed UAV system achieves meter-level accuracy and reconstructs the environment with dense point cloud.
Indoor localization using magnetic and inertial sensors has shown its effectiveness due to the pervasiveness of magnetic fields and independence from external infrastructure. However, when massive crowdsourced data are obtained in mass-market applications, new issues occur and degrade the magnetic matching (MM) performance. To alleviate these issues, an orientation-aided stochastic MM method is presented. First, orientation-aided magnetic fingerprints are applied to effectively improve fingerprint fidelity when massive data are used. Second, this paper characterizes crowdsourced magnetic fingerprints and reveals that the stochastic magnetic components may have different histograms, such as the symmetric, left-skewed, bimodal, and irregular ones. Based on such outcome, Gaussian distribution and histogram model based stochastic MM methods are presented to mitigate the degradation from stochastic magnetic components, which have not been involved in the existing MM methods such as nearest neighbor and dynamic time warping. Compared to the deterministic MM methods, this research provides a deeper insight into the use of magnetic fingerprints and more accurate MM and MM/dead-reckoning integrated localization solutions.