Active Perception for Inertial-Aided Systems

Inertial Measurement Units (IMUs) are proprioceptive sensors that measure angular velocities and linear accelerations which are integrated to estimate the relative pose of vehicles in motion. Estimation algorithms used with IMUs need to account for sensor noises which corrupt the measurements, and characterize IMU biases. This thesis proposes Informative Path Planning (IPP) frameworks that actively maximize information gain in inertial-aided perception tasks of extrinsic calibration, localization and mapping. Firstly, we propose an algorithm to generate continuous and differentiable paths based on Gaussian Process (GP) regression and Linear operators, that allows embedding of constraints in the first and second derivative spaces (velocity and acceleration measurements) in the position trajectory. These trajectories are used within an IPP algorithm that prioritizes convergence of IMU biases to improve localization accuracy. Secondly, we use IPP to find the admissible continuous and differentiable path that produces the most accurate calibration for lidar-inertial systems in an unknown environment within a given time budget. Finally, we present an active mapping framework that uses point and plane features within a Visual-Inertial Odometry framework. The continuous planner based on GP regression is used to actively choose the most informative path that will maximize information gain for our mapping objective.

Author dc:creator

• Usayiwevu, Mitchell

Statement dc:rights

• info:eu-repo/semantics/openAccess
• The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
• © 2022 Mitchell Usayiwevu
• au.edu.uts.lib/cph

Language dc:language.iso

en_US

Handle dc:identifier.uri

http://hdl.handle.net/10453/172457

OAI identifier oai:identifier

oai:opus.lib.uts.edu.au:10453/172457