Year

2022

Degree Name

Doctor of Philosophy

Department

School of Electrical, Computer and Telecommunications Engineering

Abstract

This dissertation explores distributed algorithms for calibration, localisation, and mapping in the context of a multi-robot network equipped with cameras and onboard processing, comparing against centralised alternatives where all data is transmitted to a singular external node on which processing occurs. With the rise of large-scale camera networks, and as low-cost on-board processing becomes increasingly feasible in robotics networks, distributed algorithms are becoming important for robustness and scalability. Standard solutions to multi-camera computer vision require the data from all nodes to be processed at a central node which represents a significant single point of failure and incurs infeasible communication costs. Distributed solutions solve these issues by spreading the work over the entire network, operating only on local calculations and direct communication with nearby neighbours.

This research considers a framework for a distributed robotic vision platform for calibration, localisation, mapping tasks where three main stages are identified: an initialisation stage where calibration and localisation are performed in a distributed manner, a local tracking stage where visual odometry is performed without inter-robot communication, and a global mapping stage where global alignment and optimisation strategies are applied. In consideration of this framework, this research investigates how algorithms can be developed to produce fundamentally distributed solutions, designed to minimise computational complexity whilst maintaining excellent performance, and designed to operate effectively in the long term. Therefore, three primary objectives are sought aligning with these three stages.

FoR codes (2008)

0801 ARTIFICIAL INTELLIGENCE AND IMAGE PROCESSING

Share

COinS
 

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.