Previous Abstract Return to Session C3 Next Abstract

Session C3: Multisensor Integrated Systems and Sensor Fusion Technologies

Multisensor Navigation in Urban Environment
Andrea Della Monica, Politecnico di Torino, Italy; Laura Ruotsalainen, Finnish geospatial Research Institute, Finland; Fabio Dovis, Politecnico di Torino, Italy
Location: Windjammer

GNSS receivers are designed to provide an estimate of three fundamental components for navigation: position, velocity and time. However, GNSS signals can be highly degraded or denied in some challenging environments like urban canyons, forests, mines and indoor environments. In addition to these critical scenarios, no orientation (or attitude) information is provided by standard receivers. Thus, in order to enhance robustness and performance of navigation solutions, it is possible to rely on integration of measurements coming from other sensors with GNSS, like inertial sensors, barometers, magnetometers, cameras and other systems that can be used for positioning. The objective is to go towards seamless positioning, integrating measurements adaptively based on availability and reliability of the different systems.
The integration algorithm is the core of the data fusion procedure: normally it is an estimator that is in charge of fusing information from multiple sources, giving as output the integrated solution. In our case an Extended Kalman Filter (EKF) has been used. A Kalman Filter is a set of mathematical equations that provide an estimate of the states of a process. In our specific case the states are the position, velocity and attitude vectors of the user and the inertial sensors biases. The filter guarantees that the covariance of the estimation error is minimized and its recursive form makes it well-suited to practical implementations. The Extended Kalman Filter has been implemented on Matlab, since it is mainly based on computations with vectors and matrices.
One of the most beneficial forms of hybridisation is the integration of GNSS with inertial sensors, also known as GNSS/INS integration. Since GNSS receivers can either supply processed information (PVT solution), raw measurements (such as pseudoranges or pseudorange rates), or raw signals, we can distinguish between three different integration architectures, based on the level at which the information is fused. In our work, accelerations and angular velocities retrieved from inertial sensors are integrated directly with the PVT solution estimated by the GNSS receiver. This architecture is called loosely-coupled. Inertial sensors bring the attitude information to the system as well as increased data rate. In addition, error characteristics and vulnerabilities of inertial sensors and GNSS are complementary, making the two navigation systems ideal for integration.
The work presented in the thesis has been carried out at the Navigation and Positioning Department of the Finnish Geospatial Research Institute (FGI) in Masala (Helsinki area, Finland). In order to test the performances of the integration algorithm, an experimental campaign has been performed in the city of Helsinki considering a urban canyon scenario that includes tunnels and streets surrounded by buildings, and , thus, GNSS outages occur. The itinerary has been covered using a car equipped with many sensors and systems.
The main systems used for the campaign are: U-blox GNSS receiver; Xsens inertial sensor unit (including barometer and magnetomer); Novatel SPAN system consisting of GNSS receiver and tactical-grade IMU. The first two systems, together with the integration algorithm, are our "cheap" solution, while the Novatel SPAN is a very expensive instrument that is used to build the reference solution (ground truth) with centimeter-level accuracy.
The obtained results highlight the very positive impact of integration, especially in the parts of the itinerary in which GNSS performances are heavily degraded. In fact, looking at the horizontal components of position, the GNSS-only solution presents some relevant gaps due to signal obstruction caused by tunnels and surrounding buildings. In order to obtain a continuous solution without any interruption, the hybridisation of GNSS and inertial sensors, namely the GNSS/INS integration, has been implemented.
Another positive impact of integration concerns the estimation of the altitude, namely the vertical component of the position vector. Differently from horizontal components, altitude does not benefit from inertial sensors, since accelerations along the Z axis are too small to be perceived by cheap sensors in land vehicle navigation. Due to the fact that inertial sensors cannot help to improve this result, we have to rely on the barometer, which measures variations in air pressure that can be converted in variations in altitude. The magnitude of the error of the integrated solution is approximately 55% smaller with respect to the GNSS-only solution.
Standard GNSS receivers do not provide any orientation (or attitude) information, thus inertial sensors and magnetometer are the only systems that can provide an estimate of this parameter. In our work, attitude is expressed using Euler angles, described by the roll, pitch and yaw sequence. In particular, yaw is the most relevant component of the attitude vector, since it gives information about the orientation of the vehicle. However, inertial sensors do not provide a sufficiently accurate estimation of this parameter. A definitely better solution is obtained by integrating inertial measurements with the Earth's magnetic field measurements provided by the magnetometer. The magnitude of the error, in this case, is reduced by about 75%.
The ultimate goal of our work is to provide an accurate indoor positioning solution for underground mines, where GNSS is not available at all. The mine environment is challenging for other positioning means also, namely the minerals in surroundings disturb the magnetometer measurements and there is not enough light for using a normal camera in a vision-aided positioning manner. However, thermal cameras function well also in dark mine environments and provide means for using visual odometry for resolving the motion direction and distance.
Since the integration algorithm is characterized by a very flexible structure, the next step in our development is to block the use of GNSS and integrate a thermal camera to the system instead.
In conclusion, in this work a method to obtain a continuous fused navigation solution has been proposed. The method has been tested in urban navigation for addressing various fusion challenges. In the paper we also discuss the use of thermal vision-aiding in the mine environment and the modifications needed for the integration algorithm due to its inclusion.



Previous Abstract Return to Session C3 Next Abstract