Interactive multiple model filtering for robotic navigation and tracking applications

Glavine, Patrick Joseph (2019) Interactive multiple model filtering for robotic navigation and tracking applications. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (5MB)


The work contained in this thesis focuses on two main objectives. The first objective is to evaluate the Interactive Multiple Model (IMM) filter method for robotic applications including inertial navigation systems (INS) and computer vision tracking. The second objective is to design an experimental testbed for multi-model mobile robot state estimation research in the Intelligent Systems Laboratory (ISLAB) at Memorial University. An IMM estimator uses multiple filters that run simultaneously to produce a combined weighted estimation of an observed system’s states. The weights are functions of the likelihood of how well each individual filter matches the current behaviour exhibited by the system. The performance of IMM filtering is evaluated using two different strategies for augmenting the system’s filter banks. The first method uses multiple kinematic models (constant velocity and constant acceleration models) in a mean-shift-based computer vision tracking application. The results of this experiment indicate that the IMM improves tracking performance due to its ability to adapt to the continuously changing motion characteristics of 2D blobs in videos. The second approach uses the same kinematics for each filter; however, the process and sensor noise parameters are tuned differently for each model. This method is tested in INS applications for both an automobile and a skid-steer mobile robot (Seekur Jr). Results show that the method improves INS tracking over single model Extended Kalman Filter (EKF) designs. Furthermore, an augmented state-space model containing skid-steer instantaneous center of rotation (ICR) kinematics is presented for future testing on the Seekur Jr INS. The experimental testbed designed in this thesis work is an operational data acquisition system developed for use with the Seekur Jr robot. The Seekur Jr platform has been Robot Operating System (ROS) enabled with access to data streams from 2D Lidar, 3D nodding Lidar, inertial measurement unit, digital compass, wheel encoder, onboard Global Positioning System (GPS), real-time kinematic (RTK) differential global positioning system (DGPS) ground truth, and vision sensors. The physical setup and data networking aspects of the testbed have been used for validation of an IMM filter presented in this thesis and is fully configured for future multi-model localization experiments of the ISLAB.

Item Type: Thesis (Masters)
Item ID: 14031
Additional Information: Includes bibliographical references (pages 117-126).
Keywords: Interactive Multiple Model Filter, Extended Kalman Filter, Mobile Robots, State Estimation, Computer Vision Tracking
Department(s): Engineering and Applied Science, Faculty of
Date: October 2019
Date Type: Submission
Library of Congress Subject Heading: Kalman filtering; Observers (Control theory); Inertial navigation--Mathematical models; Robot vision--Mathematical models

Actions (login required)

View Item View Item


Downloads per month over the past year

View more statistics