{"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/learning-from-the-hindsight-plan-episodic-mpc","title":"Learning from the Hindsight Plan -- Episodic MPC Improvement","arxiv_id":"1609.09001","date":"2016-09-28","proceeding":null,"authors":["Aviv Tamar","Garrett Thomas","Tianhao Zhang","Sergey Levine","Pieter Abbeel"],"abstract":"Model predictive control (MPC) is a popular control method that has proved\neffective for robotics, among other fields. MPC performs re-planning at every\ntime step. Re-planning is done with a limited horizon per computational and\nreal-time constraints and often also for robustness to potential model errors.\nHowever, the limited horizon leads to suboptimal performance. In this work, we\nconsider the iterative learning setting, where the same task can be repeated\nseveral times, and propose a policy improvement scheme for MPC. The main idea\nis that between executions we can, offline, run MPC with a longer horizon,\nresulting in a hindsight plan. To bring the next real-world execution closer to\nthe hindsight plan, our approach learns to re-shape the original cost function\nwith the goal of satisfying the following property: short horizon planning (as\nrealistic during real executions) with respect to the shaped cost should result\nin mimicking the hindsight plan. This effectively consolidates long-term\nreasoning into the short-horizon planning. We empirically evaluate our approach\nin contact-rich manipulation tasks both in simulated and real environments,\nsuch as peg insertion by a real PR2 robot.","url_abs":"http://arxiv.org/abs/1609.09001v2","url_pdf":"http://arxiv.org/pdf/1609.09001v2.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":"learning-from-the-hindsight-plan-episodic-mpc","repo_url":"https://github.com/zuoxingdong/VIN_PyTorch_Visdom","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"contact-rich-manipulation","task_name":"Contact-rich Manipulation"},{"task_slug":"model-predictive-control","task_name":"Model Predictive Control"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1609.09001","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}