Shaped-based IMU/Camera Tightly Coupled Object-level SLAM using Rao-Blackwellized Particle Filtering

Ilyar Asl Sabbaghian Hokmabadi

Shaped-based IMU/Camera Tightly Coupled Object-level SLAM using Rao-Blackwellized Particle Filtering

dc.contributor.advisor	El-Sheimy, Naser
dc.contributor.author	Ilyar Asl Sabbaghian Hokmabadi
dc.contributor.committeemember	Lichti, Derek
dc.contributor.committeemember	Shahbazi, Mozhdeh
dc.date	2023-11
dc.date.accessioned	2023-10-10T15:25:28Z
dc.date.available	2023-10-10T15:25:28Z
dc.date.issued	2023-10-04
dc.description.abstract	Simultaneous Localization and Mapping (SLAM) is a decades-old problem. The classical solution to this problem utilizes entities such as feature points that cannot facilitate the interactions between a robot and its environment (e.g., grabbing objects). Recent advances in deep learning have paved the way to accurately detect objects in the image under various illumination conditions and occlusions. This led to the emergence of object-level solutions to the SLAM problem. Current object-level methods depend on an initial solution using classical approaches and assume that errors are Gaussian. This research develops a standalone solution to object-level SLAM that integrates the data from a monocular camera and an IMU (available in low-end devices) using Rao Blackwellized Particle Filter (RBPF). RBPF does not assume Gaussian distribution for the error; thus, it can handle a variety of scenarios (such as when a symmetrical object with pose ambiguities is encountered). The developed method utilizes shape instead of texture; therefore, texture-less objects can be incorporated into the solution. In the particle weighing process, a new method is developed that utilizes the Intersection over the Union (IoU) area of the observed and projected boundaries of the object that does not require point-to-point correspondence. Thus, it is not prone to false data correspondences. Landmark initialization is another important challenge for object-level SLAM. In the state-of-the-art delayed initialization, the trajectory estimation only relies on the motion model provided by IMU mechanization (during the initialization), leading to large errors. In this thesis, two novel undelayed initializations are developed. One relies only on a monocular camera and IMU, and the other utilizes an ultrasonic rangefinder as well. The developed object-level SLAM is tested using wheeled robots and handheld devices, and an error (in the position) of 4.1 to 13.1 cm (0.005 to 0.028 of the total path length) has been obtained through extensive experiments using only a single object. These experiments are conducted in different indoor environments under different conditions (e.g. illumination). Further, it is shown that undelayed initialization using an ultrasonic sensor can reduce the algorithm's runtime by half.
dc.identifier.citation	Asl Sabbaghian Hokmabadi, I. (2023). Shaped-based IMU/camera tightly coupled object-level SLAM using Rao-Blackwellized particle filtering (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.
dc.identifier.uri	https://hdl.handle.net/1880/117351
dc.language.iso	en
dc.publisher.faculty	Graduate Studies
dc.publisher.institution	University of Calgary
dc.rights	University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.
dc.subject	Simultaneous Localization and Mapping (SLAM)
dc.subject	Shape
dc.subject	Deep Learning-based Object Segmentation
dc.subject	Extrinsic Calibration
dc.subject	Ultrasonic Rangefinder
dc.subject	Monocular Camera
dc.subject	Inertial Measurement Unit
dc.subject	Object-level
dc.subject	Particle Filter
dc.subject	RBPF
dc.subject	Contour-based
dc.subject	Pose Estimation
dc.subject	Symmetrical Objects
dc.subject	Floor Segmentation
dc.subject	Indoor
dc.subject.classification	Education--Mathematics
dc.subject.classification	Education--Tests and Measurements
dc.title	Shaped-based IMU/Camera Tightly Coupled Object-level SLAM using Rao-Blackwellized Particle Filtering
dc.type	doctoral thesis
thesis.degree.discipline	Engineering – Geomatics
thesis.degree.grantor	University of Calgary
thesis.degree.name	Doctor of Philosophy (PhD)
ucalgary.thesis.accesssetbystudent	I do not require a thesis withhold – my thesis will have open access and can be viewed and downloaded publicly as soon as possible.