Imitation Learning for Vision-based Lane Keeping Assistance

TL;DR

This paper explores using imitation learning with CNNs trained on real human driving data to develop a vision-based lane keeping system that performs well in simulation, demonstrating robustness and effective lane positioning.

Contribution

It introduces a CNN-based imitation learning approach for lane keeping using real-world grayscale images, showing promising results in simulation environments.

Findings

Successful learning of lane keeping policy from real human data

Demonstrated robustness to domain changes in simulation

Achieved good lane positioning and smooth driving trajectories

Abstract

This paper aims to investigate direct imitation learning from human drivers for the task of lane keeping assistance in highway and country roads using grayscale images from a single front view camera. The employed method utilizes convolutional neural networks (CNN) to act as a policy that is driving a vehicle. The policy is successfully learned via imitation learning using real-world data collected from human drivers and is evaluated in closed-loop simulated environments, demonstrating good driving behaviour and a robustness for domain changes. Evaluation is based on two proposed performance metrics measuring how well the vehicle is positioned in a lane and the smoothness of the driven trajectory.