Multi-Rate Unscented Kalman Filtering for Pose Estimation

Multi-Rate Unscented Kalman Filtering for Pose Estimation: Using a car-like vehicle-platform

de Vries, Maarten (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft OLD Biorobotics)

Mazo, M. (mentor)
Seiffers, John (graduation committee)

Delft University of Technology

Mechanical Engineering | Biomechanical Design - BioRobotics

2019-03-11

Pose estimation through fusion of GNSS with secondary sensors has long been an established field. With the developments surrounding autonomous navigation over the past decade this topic has gained extra importance. In the current literature GNSS based pose and localisation is often improved through fusion with either a IMU}or VS with the goal of improving on stand-alone GNSS localisation results as well as dealing with GNSS outages. In this thesis however, all three of these sensors will be fused together using a cascade of a IMU orientation filter and a Multi-Rate UKF. This filter structure is evaluated using simulations and real-world data obtained using a created vehicle-platform. The simulated results indicate that using a Multi-Rate Unscented Kalman Filter for pose estimation is promising as the filter, when configured properly, outperforms stand-alone GNSS receivers for pose estimation. However, the real-world experiments show that the used sensors lack accuracy and precision to obtain satisfactory results.

GNSS
Kalman Filter
Unscented Kalman
Pose estimation
sensor fusion
Multi-Rate
IMU sensor fusion
Vehicle-sensor
Protocol

http://resolver.tudelft.nl/uuid:e5d2506a-df25-4187-8ffa-10ffd2c819c6

Student theses

master thesis

© 2019 Maarten de Vries