Autonomous driving in structured and unstructured environments

Abstract

Autonomous vehicles are being used increasingly often for a range of tasks, including automated highway driving and automated parking. Such systems are typically either specialized for structured environments and depend entirely on such structure being present in their surroundings, or are specialized for unstructured environments and ignore any structure that may exist. In this paper, we present a hybrid autonomous system that recognizes and exploits structure in the environment in the form of driving lanes, yet also navigates successfully when no such information is present. We believe such an approach is more flexible and more robust than either of its sub-components alone. We demonstrate the effectiveness of our system on both marked roads and unmarked lots.