System Design, Modelling, and Control for an Off-Road Autonomous Ground Vehicle

Abstract

Autonomy in passenger road vehicles has long been a goal for many research groups and companies, and there has been a significant amount of focus on achieving this endeavour. A lesser focused upon area is the task of precise autonomous driving in off-road environments, where widely varying terrain geometry & material as well as vehicle limitations impede modelling and control.
This thesis explores the process of designing and retrofitting a drive by wire system to an off-road passenger utility terrain vehicle, followed by modelling and control of the vehicle. In particular, we explore tractable methods of vehicle modelling in conditions where the internal combustion engine powered vehicle has difficulty following control inputs, such as low speed over rough terrain. We present a hybrid vehicle model which combines conventional actuator and motion models with a recurrent neural network learning residuals for the conventional models' predictions. This approach allows us to use simpler conventional model formulations while leveraging machine learning methods to correct approximations and assumptions using empirical data. Finally we present a method to plan kinematically feasible paths using Euler spirals to ensure curvature continuity, making use of power series approximations to reduce the computational cost of our overall solution.