The full text of this item is not available at this time because the student has placed this item under an embargo for a period of time. The Libraries are not authorized to provide a copy of this work during the embargo period, even for Texas A&M users with NetID.
Planning and Vision Based Methods for Autonomous Vehicles
Abstract
Autonomous vehicles(AVs) have the potential to revolutionize how we ultimately perceive modern transportation. Many current car models already feature advanced driver-assist systems (ADAS), such as adaptive cruise control (ACC), Lane Departure Warning (LDW), Lane Keep As-sist (LKA), and parking assist systems that allow cars to steer themselves. Companies like Waymo are working towards achieving Level 5 autonomy, allowing vehicles to drive on existing roads and navigate various environmental conditions with little human input. However, the adoption of AVs has been slow. Studies have cited missing research as a barrier that has prevented a more widespread deployment of AVs. Through this work, I hope to address some of these research gaps in motion planning and perception of AV and contribute to the advancement of AV technology.
Chapter 2 explores mission planning approaches for a team of motion-constrained vehicles that need to visit a set of targets. The problem is formally defined as the Minmax Dubins Generalized multiple Travelling Salesman Problem (MD-GmTSP), and three main approaches for solving it are presented: 1) an Optimization approach using a Mixed Integer Programming formulation solved on a CPLEX Solver, 2) a heuristics-based approach using Variable Neighborhood search, and 3) policy modeling using Reinforcement learning. The resulting output is a global mission plan that consists of a target visitation order and feasible orientations at each target for a team of curvature-constrained vehicles. Chapter 3 investigates the problem of finding a collision-free Curvature-constrained Shortest Path (CSP) for vehicles with a minimum turning radius constraint in the presence of obstacles. The main focus of this chapter is to provide optimality guarantees for the CSP between two vehicle configurations. A novel method is presented to find tighter lower bounds (a-posteriori guarantees) for the CSP and, as a result, more accurately estimate the quality of the feasible solution. Chapter 4 investigates how lane marking can be reliably detected to assist in the safe execution of motion plans developed by the motion planning stack. An in-depth review of factors like lane marking characteristics and marking quality is conducted to evaluate the road’s quality of lane marking features. A benchmark system is proposed that can be used to evaluate the dependence of lane detection systems on lane infrastructure. Datasets incorporating pavement marking material characteristics into the lane detection framework are developed, and a systems approach is presented to correlate the lane detection algorithm performance to environmental factors, lane marking types, color, material, and the retroreflectivity of pavement markings.
Citation
Nayak, Abhishek (2023). Planning and Vision Based Methods for Autonomous Vehicles. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /198865.