|Ph.D Student||Klein Itzik|
|Subject||Mitigating INS Drift by Software Solutions|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Tomer Toledo|
|Professor Sagi Filin|
|Full Thesis text|
One of the most popular ways to perform navigation is using Inertial Navigation Systems (INS). However, INS solution degrades with time as its sensors measurements contain noise, which permeates into the navigation equations. To circumvent that drift, INS is regularly fused with other sensors or data such as GPS.
In this research we explore the case of land navigation in an urban environment. There, the vehicle is equipped with a low-cost MEMS INS/GPS unit and experiences complete or partial GPS outages for short time periods. The aim of this research is to find means to mitigate the INS drift in these situations. Specifically, we explore the possibilities of employing software solutions for INS fusion, instead of traditional hardware aiding.
Several numerical methods have been proposed in the literature (e.g. covariance analysis) in order to evaluate the aided INS performance in early stages of the design and system specification. However, due to the dependence of the INS error-state model on time and vehicle dynamics, no closed-form solution exists to evaluate the aided INS navigation performance.
We derive herein analytical solutions to evaluate the aided INS performance for short time periods. These closed-form solutions suppress the need for numerical evaluation of the aided INS performance and bring insight into the effect of the various parameters involved in such fusion.
In cases of complete GPS outages, incorporation of vehicle constrains into the estimation process is proposed. This approach only requires software adjustments in the navigation system. We derive several new vehicle constraints. To further improve the effect of these constraints two approaches are proposed: 1) implementing them into the filter system and measurement models and 2) implementing them with multiple model estimation algorithms. Both approaches show substantial improvement in reducing INS drift.
Finally, in cases of partial GPS outages, we propose a methodology to facilitate the use of the popular loosely coupled approach for INS/GPS integration. Its main disadvantage is that it is unable to provide measurement updates during periods of partial GPS availability. We use fictitious GPS satellites in order to construct a set of at least four, true and fictitious, GPS satellite and thus enable the loosely coupled approach
To demonstrate the contribution of these software solutions, several field experiments were conducted with a MEMS INS/GPS unit while driving in urban environments. The results show that the proposed approaches provide substantially lower navigation errors compared to the standalone INS.