Previous Abstract Return to Session C3 Next Abstract

Session C3: Multisensor Integrated Systems and Sensor Fusion Technologies

Modular Tightly Coupled System with Clock Error Estimation for Multi-GNSS for Road Vehicle Applications in Urban Scenarios
Johanna Rieke, Martin Escher and Ulf Bestmann, Institute of Flight Guidance, Germany
Location: Windjammer

In the past few years lots of efforts were made to realize advanced driver assistance systems (ADAS) and autonomous driving. A key feature of vehicle automation is the positioning. Especially the enhancement of accuracy and integrity of the position are important for safety critical applications. To evaluate the overall performance of a localization system the worst conditions have to be considered. In urban areas the reception of GNSS signals is limited due to obstacles within the line-of-sight to the GNSS satellites. The number of available satellite signals with an unblocked line of sight is decreased for instance by buildings, trees, road signs or other road users. The satellite navigation is also an essential part for lots of positioning solutions including inertial navigation or odometer sensors based systems.
The Institute of Flights Guidance of Technische Universität Braunschweig is doing research in the area of fusion algorithms for tightly coupled GNSS based navigation solution. The results of former studies show smaller errors in tightly coupled systems compared to common navigation solutions like loosely coupled systems. The limitation of coupled positioning algorithm appears in urban areas with a poor reception of GNSS signals. To compensate the problem of GNSS disturbed reception a modular update algorithm was developed. A clear improvement of position accuracy in road vehicle applications was achieved. This computation uses only the data of GPS system combined with IMU data and odometer information. An enhancement of the existing coupling algorithm is the integration of other GNS systems like GLONASS, Galileo and BeiDou to increase the availability of GNSS updates and to investigate the potential of position accuracy.
Former publications showed common clock error models of the receiver clock are insufficient concerning urban areas with disturbed reception. The implementation of a new universal clock model estimating parameter during run time contains difficulties. Therefore the approach of modular tightly coupled system is introduced. It activates different parts of the update depending on the number of satellites. The main feature is the clock drift estimation when receiving less than four satellite signals during a certain time slot.
The challenge of using multi-GNSS originates in four different GNSS architectures. To realize a calculation of position the received data has to be converted to one basic system concerning coordinates and time base. Each system has its own reference time. Therefor the inter system bias and drift have to be considered as well as the estimated receiver clock error. Studies with broadcast inter system bias have temporary significant deviations in positioning solutions. In the presented approach the clock error differences between the systems are compensated by using four parameters of receiver clock error, one for each system. Under the assumption that the clock drift error between the systems is very small, only one parameter for receiver clock drift estimation is initialized.
In combination with the modular update satellite signals of different systems complement one another to different kinds of system state updates.
The measurement setup is installed on a test vehicle of the institute. Raw measurements of NovAtel OEM6 GNSS receiver combined with a high end inertial navigation system are processed with the NovAtel Waypoint software "Inertial Explorer" to get high precision position information as reference system. The measurement instrumentation of the test vehicle consists of different kind of high and low cost receivers, a low cost IMU (ADIS 16375) and the on-board odometer of test vehicle. First results demonstrate refinement of altitude and velocity error.



Previous Abstract Return to Session C3 Next Abstract