Optimization-Based Sensor Fusion of GNSS and IMU Using a Moving Horizon Approach.

Fabian Girrbach,Giovanni Bellusci,Moritz Diehl,Jeroen Hol

doi:10.3390/s17051159

Fabian Girrbach, Giovanni Bellusci + Show 2 more

Open Access

https://doi.org/10.3390/s17051159

Copy DOI

Abstract

The rise of autonomous systems operating close to humans imposes new challenges in terms of robustness and precision on the estimation and control algorithms. Approaches based on nonlinear optimization, such as moving horizon estimation, have been shown to improve the accuracy of the estimated solution compared to traditional filter techniques. This paper introduces an optimization-based framework for multi-sensor fusion following a moving horizon scheme. The framework is applied to the often occurring estimation problem of motion tracking by fusing measurements of a global navigation satellite system receiver and an inertial measurement unit. The resulting algorithm is used to estimate position, velocity, and orientation of a maneuvering airplane and is evaluated against an accurate reference trajectory. A detailed study of the influence of the horizon length on the quality of the solution is presented and evaluated against filter-like and batch solutions of the problem. The versatile configuration possibilities of the framework are finally used to analyze the estimated solutions at different evaluation times exposing a nearly linear behavior of the sensor fusion problem.

Highlights

During the last few years, fully autonomous systems have been a highly active research field, which pushed product development towards the commercialization of such systems
The complete sensor setup used for data collection contains two independent sets of sensors: one for the estimation results and one to obtain the reference trajectory. Both sets use a global navigation satellite system (GNSS) receiver and an inertial measurement unit (IMU), which differ in their specifications
The presented Moving horizon estimation (MHE) approach for multi-sensor fusion allows a mathematically elegant formulation in a single optimization problem where cost functions and constraints are recursively adapted with the availability of new measurement data

Summary

Introduction

During the last few years, fully autonomous systems have been a highly active research field, which pushed product development towards the commercialization of such systems For applications such as autonomous driving and unmanned aerial vehicles (UAVs) more semi-autonomous features become available every day. The first prototypes have already proven to be feasible and have successfully delivered small sized goods, simplified inspection processes in rough terrain or harvested wind energy at previously unreachable altitudes. These recent developments allow the prediction of an increasing number of UAVs applications (see Figure 1), which will result in a more crowded airspace, representing a paradigm shift in comparison to traditional airborne applications where the airspace is heavily secured and supervised.

Objectives

Methods

Results

Conclusion