Previous Abstract Return to Session A2 Next Abstract

Session A2: Small Size or Low Cost Inertial Sensor Technologies

Loosely Coupled GPS/INS Integration with Snap to Road for Low-Cost Land Vehicle Navigation
Mohamed Lajmi Cherif, Jérôme Leclere and René Jr. Landry, University of Québec, École de Technologie Supérieure, Canada
Location: Big Sur

Nowadays, the use of navigation system is getting more and more essential. These systems can be useful in different fields; marine navigation, routing, car traffic etc,… We can name the most useful systems, Global Positioning System (GPS) and inertial navigation system (INS),…
Many projects have been conducted in this area. The objective of such projects is to enhance the performance, availability of positions and decrease errors and cost.
But, with respect to the existing works, some of them have some limitations. For example, the location with Global Positioning Systems (GPS) in severe environments such as urban canyons, tunnels, forested areas and bridges is difficult and even non-existent. The use of a complementary navigation system is necessary therefore the inertial navigation system (INS) has been chosen. The INS allows to have more information on the navigation which makes it possible to consult the location even in the severe environments but to have a good precision the latter must cost more than 100.000 $. The use of a low cost system is necessary but we encountered a problem of precision because the latter will diverge enormously in the case there is not a GPS signal. In this context, we should consider the coupe of parameters cost vs precision and try to optimize it.
Among the techniques that have been used, we can cite the GPS/INS integration which is an approach to use the GPS signal in order to correct and calibrate an inertial data. The signals produced by the INS are accurate for a small duration, because of successive integrations necessary for the variables of position and speed. However, the advantage is that they provide data at high frequency. Conversely, GPS data are precise, don't deviate with the time, but their frequency is lower and it can happen that the GPS signal is lost because of external parameters. We use GPS signals to periodically reset INS signals.
The most commonly used approach in this domain is Kalman Filter which is an infinite impulse response filter that estimates the states of a dynamic system from a series of incomplete or noisy measurements.
The first limitation that we need to consider is the fact that the Kalman filter is an approximation of the first order, which explains its relative ease of implementation and speed of calculation. However, this approximation is insufficient in terms of accuracy and robustness when else nonlinearities are too strong. Although, numerous techniques can be combined with Kalman Filter namely, Loosely Coupled, Tightly Coupled, Ultra-Tightly Coupled … based on the degree of the integration between GPS and INS systems. The mostly used approach is the Loosely Coupled. It is the easiest to use and it provides results in a short amount of time. Unlikely, the Tightly Coupled and Ultra-Tightly Coupled techniques are more complex and deserve much computation time, but they provide more accurate results.
Snap to Road (STR) is another technique that has been used in the area of system navigation. STR is a service provided by Google. It uses a set of point positions and outputs their best coordination with respect to the geometry of the road.
The major advantage of STR is to give the best approximation/projection of the most likely road a vehicle could have been following. But it also has some limitation related to the topology of the road map. It can be sometime a subject to inconsistency issues with respect to the most probable route calculation. These shortcomings can be decreased when combined with other techniques that will play the role of calibration.
In our approach, we propose to use a specific combination between the above described techniques. Thus, we can take advantage of their powerful characteristics. In order to do so, we used loosely coupling GPS/INS integration within the Extended Kalman Filter (EKF) together with STR. EKF is an improved version of Kalman filter that cope with non-linear behavior (also called nonlinear Kalman Filter) which will be a best architecture to fit with our targeted systems.
In this article, we present a new method of coupling integration between INS and GPS which will use STR system as an asset to guarantee the viability of the positions and to provide small range error.
Our proposed solution is implemented in order to have more precision in navigation information and to ensure accurate and robust results even in severe environments.
The general idea of our work is to use the data (navigation points, IMUs) provided by INS and GPS parameters as an input to the Kalman Filter where the process of adjustment will be later conducted. Later, the output will be fed to STR to provide a linear solution of the route. The use of STR in our method is covering different scenarios based on whether we have outage or not.
In order to achieve the proposed objective, tow system are used. The first one is the IMU-CPT Tactical Grade Fiber Optic Gyros, which was integrated with the Novatel ProPak6™ Triple-Frequency GNSS Receiver. It is composed of Fiber Optic Gyros (FOG) and Micro Electromechanical Systems (MEMS) accelerometers. FOGs offer exceptionally long life and stable performance compared with other similar gyro technologies. Using the data form Novatel system in the inertial explorer program to generate a tightly coupled INS/GPS navigation solution, which is used as a reference to compare the performance and the effectiveness of the proposed methods when applied to MEMS based sensors. The second system is the micro-iBB VTADS which was developed in our laboratories using the low-cost equipment for IMU STMicroelectronics; L3GD20 MEMS motion sensor: 3-axis digital gyroscope and LSM303DLHC 3D accelerometer and 3D magnetometer module that utilize the suggested methods. The experimental test setup, was built as platform mounts for the car. Serval road trajectory tests were carried out using the above-described setup.
In order to show the usefulness of our method, we performed different test operations in different environments. Compared to the following techniques,
- EKF: Extended Kalman Filter
- GPS
- STR-GPS: GPS technique combined to STR
our results are more accurate, more precise and in most of the cases, they follow the STR reference behavior (Ref-STR) which is a slightly drifted version of the “ideal” behavior (Ref). The other approaches provide in some points a high level of divergence from the Ref-STR route which leads to inaccurate results.
INS-STR technique has been implemented. It consists of a combination of INS systems and Snap to Road technique. We are intended to compare its results to ours in different scenarios and show how our method is leading in this context.
Recently, many projects targeted the improvement of navigation systems. We proposed to use the STR technique together with INS and GPS in order to analyze the adjusted behavior of routing within harsh environments. So far, the output results show that our method provides more accurate results a small amount of error. Our intention is to use calibration algorithms (that we will develop) in order to improve further our obtained results and to cope with the shortcomings of the Google maps. The main idea is to provide results which are more like the reference route (Ref) than the STR-Ref one.



Previous Abstract Return to Session A2 Next Abstract