Previous Abstract Return to Session D3 Next Abstract

Session D3: GNSS Augmentation and Robustness for Autonomous Navigation

Evaluation of Dual-Frequency Multi-Constellation PPP/INS using QZSS Augmentation Data
Cheng-Wei Wang, Shau-Shiun Jan, Department of Aeronautics and Astronautics, National Cheng Kung University
Alternate Number 1

Since highly demands for autonomous driving applications, accurate navigation solution is required. Lane-level navigation should be satisfied to enable the robust vehicle guidance. GNSS provides user position and the solution accuracy is independent of trajectory and time. However, GNSS is vulnerable to environmental factors in urban area. Therefore, GNSS/INS integration system is well developed, which mitigates the disadvantages of the two systems. Accurate acceleration and attitude can be measured, and external information is significant for road users. Tightly couple scheme is chosen as the fusion framework to keep the solution propagate even in GNSS signal blocked condition or multipath environment. For GNSS solution, PPP is applied because of its high positioning accuracy. In recent years, faster convergence time PPP is realizable with INS and SBAS augmentation signal aided. Quasi-Zenith Satellite System (QZSS) is a regional navigation satellite system which covering Eastern Asia and Oceania region. The augmentation data, also called the Multi-GNSS Advanced Demonstration tool for Orbit and Clock Analysis (MADOCA) product is developed by JAXA. The aim of MADOCA orbit products is to provide satellite orbit within 10 cm accuracy in real-time and offline. Therefore, MADOCA real-time products in RTCM State-Space Representation (SSR) format are conducted in this research. The augmentation data is broadcasted using L6E signal which correction messages such as satellite orbit and clock correction information are included. Moreover, wider service area makes it possible to conduct in Taiwan. In this paper, performance of MADOCA-PPP in Taiwan combining broadcast ephemeris with augmentation data will be evaluated. Multi-constellation GNSS such as GPS, GLONASS and QZSS are considered because MADOCA product offers the correction messages about the three constellations and sound satellite geometric distribution makes the navigation solution more reliable. In addition, the convergence time may reduce with more satellites available. However, ionospheric delay/advance may cause significant impact on ranging measurement errors. The dual-frequency measurement combination is thus considered to eliminate the ionospheric ranging errors.
The concept of PPP has been developing for a long time. PPP is usually used for geodetic surveying, sea level monitoring with buoy because centimeter-level positioning accuracy is needed. Researchers are focusing on reducing the horizontal positioning error with static reference station. The accuracy can be 10 cm or even 1 cm with the precise satellite orbit and clock product provided by International GNSS Service (IGS) network. Taking Multi-GNSS Experiment (MGEX) project for example, the GNSS signal-in-space ranging error (SISRE) is approximately 5 cm or better. Recently, instant PPP becomes a popular issue. Wide area PPP is further available on agricultural and ship positioning applications with advantage of no base station required, region based centimeter-level accuracy and reducing convergence time. QZSS transmits correction message to realize the concept in Eastern Asia. The MADOCA-LEX signal corrects satellite orbit error to less than 10 cm in real-time, which is competitive with IGS product. Decimeter-level accuracy is reachable with kinematic PPP using corrected GPS information. For lane-level navigation, the accuracy should be less than 1.5 m in 3D direction. PPP with INS aiding enhance the navigation performance to meet the requirement even in challenging area. The most common strategies to fuse the solution are loosely couple and tightly couple schemes. Loosely couple scheme might suffer from the rapid drift solution in GNSS outage condition. In addition, navigation can not propagate while less than four satellites are available. The tightly couple scheme combines GNSS and INS measurement for state propagation, which is helpful for bridging gap the problems.
As mentioned above, the tightly couple GNSS/INS provides better performance than GNSS only solution. The MADOCA-PPP without INS is also a significant issue to explore in this paper. Benefit from code phase and carrier phase measurements are included in the PPP model with several error source correction, and it is characterized with high positioning accuracy. Kinematic PPP with shorter convergence time using SBAS augmentation data is no longer leaving researchers in the dust. The augmentation message in RTCM format is downloaded from JAXA website. First, the binary file will be decoded to obtain correction information about the three constellations. The parameters are going to correct satellite orbit and clock error after deriving the orbit from broadcast ephemeris. Dual-frequency multi-constellation (DFMC) MADOCA-PPP is solved using Extended Kalman Filter (EKF). Static MADOCA-PPP is evaluated at first. Performance and analysis of different scenarios will be considered, including conventional PPP without MADOCA, MADOCA-PPP with single-frequency single-constellation, single-frequency multi-constellation, dual-frequency multi-constellation case. The convergence time improvement, root mean square error (RMSE) and standard deviation (STD) are compared across these scenarios. Multiple ranging error models such as antenna phase correction, differential code bias, tropospheric delay, ionospheric delay/advance and Earth solid tide correction, are also considered in PPP. Besides, satellite side error model corrected by correction messages will be discussed. The standalone PPP might pose severe problems in urban or sheltered area and the navigation solution can not propagate continuously because GNSS signals are suffered from environmental factors. Thus, multi-sensor navigation has attracted more attention since the rapid growth of autonomous driving cars. Multiple sensors may be integrated to strengthen the navigation availability which meet the requirement for road users and overcome challenging environments when GNSS signals are blocked. In this research, kinematic MADOCA-PPP is integrated with tactical grade IMU using tightly couple scheme, which makes the algorithm robustly. For standalone kinematic MADOCAPPP, receiver velocity and acceleration are added in static PPP state model. For integration scheme, the state model includes position error, attitude error, velocity error, gyro and accelerometer biases, receiver clock bias and drift, tropospheric delay, carrier ambiguities. Applying EKF as the main framework to fuse raw GNSS observables and IMU measurement, the poor navigation performance in challenging area could be mitigated. The experiment will be conducted in several scenarios, including kinematic PPP, kinematic MADOCA-PPP and tightly couple MADOCA-PPP/INS in single/dual-frequency and single/multi-constellation case. STD, RMSE and convergence time will also be discussed. All the algorithm is processing in post-mission but applying the real-time satellite orbit and clock correction product.
The DFMC PPP is proposed with QZSS MADOCA SSR real-time product aiding to make the analysis approach navigation solution in real world. Satellite orbit and clock error from broadcast ephemeris can be corrected by the transmitting L6E signal. For post-mission analysis, MADOCA products are acquired via Internet. The convergence time is defined as horizontal and vertical error lower than 40-centimeter level constantly. Faster convergence time is anticipated to achieve compared to conventional PPP without augmentation message. A GNSS receiver and antenna is able to process the algorithm. Reference station is not required to conduct differential corrections. However, horizontal and vertical direction bias might not meet the requirement for lane-level navigation. An INS integrated system can not only improve the performance in 3D direction, but also the navigation solution can still propagate even in GNSS outage condition. The integrated system works in all environments with higher data output rate than GNSS and provides short-term accuracy. Combining long-term accuracy of GNSS solution, high positioning accuracy, high acceleration measurement rate and precise attitude determination are available. The experiments are conducted in open sky and urban area to evaluate the convergence time difference and positioning accuracy between the two scenarios. GNSS signal blocked situation is also considered to verify if the MADOCA-PPP/INS integration scheme can still work or drift far away. Last but not least, convergence time of DFMC PPP will be compared before and after INS aiding to evaluate whether the coupling scheme is capable of shortening convergence time. Decimeter-level positioning accuracy is anticipated to achieve with the integration scheme.



Previous Abstract Return to Session D3 Next Abstract