{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/query-efficient-imitation-learning-for-end-to","title":"Query-Efficient Imitation Learning for End-to-End Autonomous Driving","arxiv_id":"1605.06450","date":"2016-05-20","proceeding":null,"authors":["Jiakai Zhang","Kyunghyun Cho"],"abstract":"One way to approach end-to-end autonomous driving is to learn a policy\nfunction that maps from a sensory input, such as an image frame from a\nfront-facing camera, to a driving action, by imitating an expert driver, or a\nreference policy. This can be done by supervised learning, where a policy\nfunction is tuned to minimize the difference between the predicted and\nground-truth actions. A policy function trained in this way however is known to\nsuffer from unexpected behaviours due to the mismatch between the states\nreachable by the reference policy and trained policy functions. More advanced\nalgorithms for imitation learning, such as DAgger, addresses this issue by\niteratively collecting training examples from both reference and trained\npolicies. These algorithms often requires a large number of queries to a\nreference policy, which is undesirable as the reference policy is often\nexpensive. In this paper, we propose an extension of the DAgger, called\nSafeDAgger, that is query-efficient and more suitable for end-to-end autonomous\ndriving. We evaluate the proposed SafeDAgger in a car racing simulator and show\nthat it indeed requires less queries to a reference policy. We observe a\nsignificant speed up in convergence, which we conjecture to be due to the\neffect of automated curriculum learning.","url_abs":"http://arxiv.org/abs/1605.06450v1","url_pdf":"http://arxiv.org/pdf/1605.06450v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"query-efficient-imitation-learning-for-end-to","repo_url":"https://github.com/mbhenaff/EEN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"autonomous-driving","task_name":"Autonomous Driving"},{"task_slug":"car-racing","task_name":"Car Racing"},{"task_slug":"imitation-learning","task_name":"Imitation Learning"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1605.06450","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}