الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 208 |  |  |
| | | from | 208 | | |
Abstract
In this chapter, summary of the proposed algorithms and studies is presented. In addition, the chapter discusses the con- clusions of each study, in addition to the recommendations re- lated to the concerned issues.
6.1 Summary
The continuous development of ITS applications and the de- ployment of the developed technologies depend on the existence of reliable and robust positioning systems that can be used in harsh urban and suburban environments. In these environments, integrating different navigation systems is essential to fulfil the positioning accuracy requirements. Employing GNSS in position- ing is crucial to provide absolute positions and aid the solutions when GNSS is integrated with other navigation systems. How- ever, in urban and suburban areas, the performance of GNSS is hampered by the harsh signal environment.
The research in this dissertation can be divided into three
main studies. The first study is concerned with developing and testing an algorithm to detect and exclude the existing GNSS NLOS signals using LiDAR sensors. In the second study, an al- gorithm is developed and tested to integrate GNSS observations with 3D building models constructed from VGI to detect the NLOS satellites and exclude the corresponding observations. Fi- nally, the third study aims to exploit the integration of LiDAR, gyroscope, and odometer sensors to develop a real-time algorithm that is capable of providing continuous and reliable navigational solutions during GNSS outages.
In the first research study, a new algorithm is proposed ex- ploiting the integration of GNSS and LiDAR sensors to predict the visibility of satellites and avoid the NLOS effect of GNSS observations. The proposed algorithm is divided into four main stages. The azimuth and elevation angles of all observed satel- lites are determined in the first stage. In the second stage, the point cloud acquired from the LiDAR sensor in the body frame is aligned to the local frame. The third stage aims to conduct a rigorous accuracy analysis to form the ‘search strips’ and de- termine the points included inside for further processing. In the final stage, the NLOS satellites are detected, and the correspond- ing observations are excluded. Real field data are used to test the proposed algorithm.
The second study presents another algorithm to detect the NLOS satellites and exclude the corresponding signals. It applies a novel processing strategy by combining an FDE algorithm with
3D building models constructed from VGI. Data from OSM and GE are integrated to build the 3D models. The algorithm is divided into five main stages. The first stage aims to compute the approximate position of the user’s receiver. In the second stage, the LOSs between the receiver and each satellite are represented mathematically in parametric forms. In the following stage, the sides (planes) of each building inside a specified search radius are extracted from a constructed 3D model of buildings. Then, in the fourth stage, a line-to-plane intersection algorithm is utilized with each LOS and the extracted planes to find the location of each intersection point. In the final stage, the NLOS signals are detected and excluded before deriving the positional solution. The proposed algorithm is tested using real field data to assess its detection capability.
Finally, the third study presents a novel real-time algorithm to provide reliable and continuous navigational solutions dur- ing long GNSS outages by integrating LiDAR, gyroscope, and odometer sensors. The algorithm adopts the line-based scan matching and applies conditions/constraints to adapt the tech- nique to the harsh environments in ITS applications. Moreover, FDE is performed to detect and exclude the incorrectly matched lines between epochs. A “road segmentation” process is also in- corporated into the algorithm to deal with the possible accumu- lation of angular errors and correct the azimuth of the moving vehicle. The presented algorithm is tested using real data after introducing three five-minute GNSS outages. The collected data
are processed using software developed by the researcher.
6.2 Conclusions
The conclusions of the three studies presented in this con- tribution are discussed in this section. Each of the following subsections is concerned with one study.
6.2.1 Detection of NLOS Signals Using LiDAR
After testing the proposed method and excluding the NLOS signals, the accuracy improvement is evaluated employing phase- smoothed code observations in the SPP processing mode. The results show that the accuracy improvement in the horizontal direction reaches 10.40m with a mean value of 2.16m (62.2% im- provement) throughout the epochs with detected NLOS signals. Also, the RMSE drops by 3.31m (63% improvement). By analyz- ing the improvement in the CT and AT directions, it is found that the accuracy improvement in the CT direction reaches 8.64m with a mean value of 1.70m, and in the AT direction reaches 6.88m with a mean value of 1.30m. The RMSE drops by 2.56m in the CT direction, while it drops by 2.11m in the AT direction.
6.2.2 Detection of NLOS Signals Using VGI 3D Building Models
After testing the presented algorithm and excluding the NLOS satellites, the accuracy improvement is assessed using SPP and
employing phase-smoothed code observations. The results showed that the accuracy improvement in the horizontal direction reaches 10.72m with a mean value of 0.80m over all epochs with detected NLOS signals, and the RMSE drops by 1.64m (46% improve- ment). Also, the improvement is analyzed in the CT and AT directions. It reaches 8.64m in the CT direction with a mean value of 0.63m, while it reaches 6.89m in the AT direction with a mean value of 0.46m. The RMSE drops by 1.24m in the CT direction, while it drops by 1.08m in the AT direction.
6.2.3 LiDAR-Gyroscope-Odometer Integration
The results show that the proposed integration and algorithm can achieve a very promising navigation performance, especially in dense urban environments where more line features can be extracted. Also, as more lines could be extracted and matched throughout the second and third outages, the performance was better over them. They achieved 2.25m and 1.89m mean posi- tional errors over the five minutes, while the maximum errors were 4.97m and 3.99m, respectively. These results indicate the robustness and reliability of applying the proposed algorithm dur- ing long GNSS outages, which can be exploited in many ITS applications.
6.3 Contributions
• A novel algorithm is developed for the detection of GNSS NLOS signals using LiDAR sensors. Rigorous accuracy
analysis of the existing error sources is applied in the algo- rithm.
• A new algorithm is proposed to exclude GNSS NLOS sig- nals using 3D building models constructed from VGI. GE and OSM data are combined to build the 3D models in the algorithm.
• A novel real-time algorithm is proposed for the integration of LiDAR, gyroscope, and odometer sensors. FDE is per- formed to eliminate the incorrectly matched lines between scans. A “road segmentation” process is employed to deal with the possible accumulation of angular errors.
• The developed algorithms in this dissertation were tested using software programmed by the researcher.
6.4 Recommendations and Future Starts
Following the previous conclusions of this dissertation, the following recommendations are addressed:
• The used LiDAR sensor in this research has a vertical FoV of 30o, which could influence the detection capability and the efficiency of using the proposed algorithms. However, significant improvements in the obtained positional accu- racy could be achieved. With the development and the ex- pected price DROP of LiDAR sensors, a larger vertical FoV
can be used, and the proposed algorithms can be applied in the same way with complete efficiency.
• Cheaper LiDAR sensors exist in the market nowadays. These sensors need to be tested to evaluate the performance that can be provided by applying the proposed techniques in a cheaper way.
• Regarding GNSS positioning using the proposed algorithms, the achieved accuracies are very promising to a wide range of urban applications such as way-point navigation, pay- per-use insurance, on-street parking, electronic toll collec- tion, and emergency services management. However, at few epochs, the accuracy decreased to few meters, which is not acceptable for applications like lane-keeping systems and collision-warning systems. Consequently, other positioning techniques such as PPP and PPP-AR need to be evaluated after adapting the presented algorithms and strategies for these safety-of-life applications.
• simplifying building footprints in OSM data and building heights in GE data can lead to missed detections or in- correct detections. In addition, some cities around the world are not fully covered by OSM footprint data and GE height data. Consequently, complementing and refin- ing the formed VGI models using other open-access data sources represent a rich research topic.
• In the absence of GNSS solutions, the integration of Li-
DAR, gyroscope, and odometer sensors using the proposed algorithm could provide very promising results over the 5- minute outages. In the performed 15-minute GNSS outage test, the lane-accuracy, which is required for some applica- tions like lane-keeping systems and collision-warning sys- tems, was satisfied in the first 1.5 minutes. Consequently, future works are needed to reach longer periods that sat- isfy the lane-accuracy during GNSS outages, exploiting the continuous development and price DROP of sensors.
• Online calibration strategies of odometer measurements can be developed to improve their accuracy and reliability. Im- proving the quality of these measurements can improve the provided navigation performance. In addition, integrating accelerometers with the used sensors may improve the over- all performance of the system.
• The proposed LiDAR-Gyroscope-odometer integration is employed in the uncoupled mode. However, coupling strate- gies can be employed by applying fusion techniques such as Kalman filters, particle filters, and artificial intelligence. The navigation performance can be reassessed by employ- ing such fusion algorithms.