Deep reinforcement Learning for end-to-end driving is limited by the need of complex reward engineering. Sparse rewards can circumvent this challenge but suffers from long training time and leads to sub-optimal policy. In this work, we explore full-control driving with only goal-constrained sparse reward and propose a curriculum learning approach for end-to-end driving using only navigation view maps that benefit from small virtual-to-real domain gap. To address the complexity of multiple driving policies, we learn concurrent individual policies selected at inference by a navigation system. We demonstrate the ability of our proposal to generalize on unseen road layout, and to drive significantly longer than in the training.
@article{arxiv.2103.09189,
title = {Goal-constrained Sparse Reinforcement Learning for End-to-End Driving},
author = {Pranav Agarwal and Pierre de Beaucorps and Raoul de Charette},
journal= {arXiv preprint arXiv:2103.09189},
year = {2021}
}