Date of Award
12-2022
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Automotive Engineering
Committee Chair/Advisor
Dr. Bing Li
Committee Member
Dr. Chris Paredis
Committee Member
Dr. Corina Sandu
Committee Member
Dr. Christopher McMahan
Abstract
In recent years, there has been an increased interest in implementing intelligent robotic systems in outdoor environments. Paramount to accomplishing this objective is being able to conduct successful robotic navigation in unprepared outdoor environments. This presents unique challenges in that there is a risk of catastrophic immobilization in terrain regions which, though unoccupied, cannot provide traction support for vehicle mobility. Methods for providing prior knowledge and perception of traction support is therefore an interest and focus of research.
In the advent of ever advancing machine learning models, “learn-as-you-go” approaches have emerged as topics of interest for mobility prediction. These approaches, however, do not address a priori mobility perception needs. Very recently, remote sensing science perception research has been addressing this challenge by developing and characterizing spectral relationships to terrain mechanical response properties. Current methods and research studies in this area solely rely on passive nm optical sensors which make surface observations of terrain and are not able to observe subsurface properties. Also, deep learning prediction frameworks have yet to be utilized due to data availability challenges.
This dissertation research investigated and developed models for terrain penetration resistance (soil stiffness) prediction from surface/subsurface measurements from a developed Vis-IR-RADAR platform. A terrain profiling system and a multiple view stereo labeling scheme was developed to generate a soil stiffness prediction dataset to enable deep learning-based prediction. Variable importance analysis results revealed insights related to common robotic sensing platforms for mobility prediction and the best nm and cm sensor modalities for mobility perception.
Recommended Citation
Brand, Howard, "Non-destructive Terrain Evaluation and Modeling for Off-Road Autonomy" (2022). All Dissertations. 3249.
https://open.clemson.edu/all_dissertations/3249