Optical Methods for Autonomous Navigation Systems

Abstract

Optical navigation has recently become increasingly popular as a method of guidance, navigation and control due to its unique advantages over other navigation systems. As a non-contact and non-destructive system, optical navigation has the potential to provide high precision navigation over large distances and is less susceptible to disruptions from a variety of environments. Autonomous navigation can further be enabled using optical sensors to detect objects, determine the vehicle’s state and guide it along predetermined paths. This makes optical navigation particularly advantageous in environments where GPS coverage is limited or unreliable such as indoors or in urban canyons. Moreover, optical navigation systems can offer higher resolution and accuracy than traditional non-optical systems in addition to being more compact and power-efficient. As technology advances, optical navigation will continue to be a viable and cost-effective solution for navigating challenging environments.

This dissertation presents a range of innovative optical sensing systems designed for onboard applications to enable autonomous landing and proximity operations in complex and rapidly changing environments. The technical details of the sensor algorithms and the results of their experimental assessment are discussed along with recommendations for possible enhancements for further work. The first chapter delves into the concept of utilizing the Doppler shift of a modulated optical signal to measure the translational relative speed between a signal generator and static photosensor. By comparing the frequency shift of the transmitted signal with local oscillation, quantitative values are derived while electric components of the sensor are meticulously customized to optimize the signal gain. Subsequently, the second chapter expands upon this idea and proposes multi-dimensional rate estimation. For this project, the light source is replaced from LED to laser, and the measure of change of rate of phase shift of the waveform is used instead of the Doppler shift measurement due to limitations in working range and modulation problems of LED. Chapter three introduces an integrated system which combines structured light beacons and a machine vision camera to facilitate autonomous landing by computing altitudes and orientations relative to the terrain. Additionally, an application of the same landing assistance algorithm is introduced, but with the use of a LiDAR-camera fusion sensing system for more precise and reliable results. Sophisticated computer vision techniques are widely utilized to detect a vanishing point of runway for heading angle computation in this application. Finally, the fourth chapter proposes a novel approach to sensor fusion that combines machine vision data and inertial measurements using the extended Kalman filter (EKF) and unscented Kalman filter (UKF) to estimate the relative pose. In comparison to similar 3D mapping and localization research, this project’s key contribution lies in the implementation of a multiplicative approach to the camera measurement model in the filter. Furthermore, the revised template matching algorithm applied to derive the measurement model is thoroughly discussed.