{"volume":41,"intvolume":"        41","year":"2022","file_date_updated":"2022-06-28T08:32:58Z","acknowledgement":"This work is graciously supported by the following grant agencies: FWF Lise Meitner (Grant M 3319), SNSF (Grant 200502), ERC Starting Grant (MATERIALIZABLE-715767), NSF (Grant IIS-181507).\r\n","ddc":["000"],"project":[{"name":"Perception-Aware Appearance Fabrication","_id":"eb901961-77a9-11ec-83b8-f5c883a62027","grant_number":"M03319"},{"grant_number":"715767","_id":"24F9549A-B435-11E9-9278-68D0E5697425","name":"MATERIALIZABLE: Intelligent fabrication-oriented Computational Design and Modeling","call_identifier":"H2020"}],"doi":"10.1145/3528223.3530144","author":[{"first_name":"Michael","last_name":"Piovarci","id":"62E473F4-5C99-11EA-A40E-AF823DDC885E","full_name":"Piovarci, Michael"},{"full_name":"Foshey, Michael","last_name":"Foshey","first_name":"Michael"},{"first_name":"Jie","last_name":"Xu","full_name":"Xu, Jie"},{"full_name":"Erps, Timothy","last_name":"Erps","first_name":"Timothy"},{"last_name":"Babaei","first_name":"Vahid","full_name":"Babaei, Vahid"},{"full_name":"Didyk, Piotr","first_name":"Piotr","last_name":"Didyk"},{"full_name":"Rusinkiewicz, Szymon","first_name":"Szymon","last_name":"Rusinkiewicz"},{"full_name":"Matusik, Wojciech","first_name":"Wojciech","last_name":"Matusik"},{"orcid":"0000-0001-6511-9385","first_name":"Bernd","last_name":"Bickel","id":"49876194-F248-11E8-B48F-1D18A9856A87","full_name":"Bickel, Bernd"}],"has_accepted_license":"1","publication":"ACM Transactions on Graphics","publication_identifier":{"issn":["0730-0301"],"eissn":["1557-7368"]},"external_id":{"arxiv":["2201.11819"]},"ec_funded":1,"article_type":"original","citation":{"mla":"Piovarci, Michael, et al. “Closed-Loop Control of Direct Ink Writing via Reinforcement Learning.” <i>ACM Transactions on Graphics</i>, vol. 41, no. 4, 112, Association for Computing Machinery, 2022, doi:<a href=\"https://doi.org/10.1145/3528223.3530144\">10.1145/3528223.3530144</a>.","ieee":"M. Piovarci <i>et al.</i>, “Closed-loop control of direct ink writing via reinforcement learning,” <i>ACM Transactions on Graphics</i>, vol. 41, no. 4. Association for Computing Machinery, 2022.","chicago":"Piovarci, Michael, Michael Foshey, Jie Xu, Timothy Erps, Vahid Babaei, Piotr Didyk, Szymon Rusinkiewicz, Wojciech Matusik, and Bernd Bickel. “Closed-Loop Control of Direct Ink Writing via Reinforcement Learning.” <i>ACM Transactions on Graphics</i>. Association for Computing Machinery, 2022. <a href=\"https://doi.org/10.1145/3528223.3530144\">https://doi.org/10.1145/3528223.3530144</a>.","apa":"Piovarci, M., Foshey, M., Xu, J., Erps, T., Babaei, V., Didyk, P., … Bickel, B. (2022). Closed-loop control of direct ink writing via reinforcement learning. <i>ACM Transactions on Graphics</i>. Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3528223.3530144\">https://doi.org/10.1145/3528223.3530144</a>","ista":"Piovarci M, Foshey M, Xu J, Erps T, Babaei V, Didyk P, Rusinkiewicz S, Matusik W, Bickel B. 2022. Closed-loop control of direct ink writing via reinforcement learning. ACM Transactions on Graphics. 41(4), 112.","ama":"Piovarci M, Foshey M, Xu J, et al. Closed-loop control of direct ink writing via reinforcement learning. <i>ACM Transactions on Graphics</i>. 2022;41(4). doi:<a href=\"https://doi.org/10.1145/3528223.3530144\">10.1145/3528223.3530144</a>","short":"M. Piovarci, M. Foshey, J. Xu, T. Erps, V. Babaei, P. Didyk, S. Rusinkiewicz, W. Matusik, B. Bickel, ACM Transactions on Graphics 41 (2022)."},"language":[{"iso":"eng"}],"_id":"11442","article_number":"112","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_created":"2022-06-10T06:41:47Z","oa":1,"oa_version":"Submitted Version","publisher":"Association for Computing Machinery","related_material":{"link":[{"relation":"press_release","description":"News on ISTA website","url":"https://ista.ac.at/en/news/machine-learning-3d-printing-fluids/"}]},"status":"public","quality_controlled":"1","title":"Closed-loop control of direct ink writing via reinforcement learning","day":"01","file":[{"success":1,"file_id":"11467","date_created":"2022-06-28T08:32:58Z","file_size":33994829,"date_updated":"2022-06-28T08:32:58Z","content_type":"application/pdf","creator":"dernst","access_level":"open_access","file_name":"2022_ACM_acceptedversion_Piovarci.pdf","checksum":"27f6fe41c6ff84d50445cc9b0176d45b","relation":"main_file"}],"type":"journal_article","tmp":{"image":"/images/cc_by.png","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"date_updated":"2023-05-31T12:38:21Z","article_processing_charge":"No","month":"06","issue":"4","publication_status":"published","date_published":"2022-06-01T00:00:00Z","abstract":[{"lang":"eng","text":"Enabling additive manufacturing to employ a wide range of novel, functional materials can be a major boost to this technology. However, making such materials printable requires painstaking trial-and-error by an expert operator,\r\nas they typically tend to exhibit peculiar rheological or hysteresis properties. Even in the case of successfully finding the process parameters, there is no guarantee of print-to-print consistency due to material differences between batches. These challenges make closed-loop feedback an attractive option where the process parameters are adjusted on-the-fly. There are several challenges for designing an efficient controller: the deposition parameters are complex and highly coupled, artifacts occur after long time horizons, simulating the deposition is computationally costly, and learning on hardware is intractable. In this work, we demonstrate the feasibility of learning a closed-loop control policy for additive manufacturing using reinforcement learning. We show that approximate, but efficient, numerical simulation is\r\nsufficient as long as it allows learning the behavioral patterns of deposition that translate to real-world experiences. In combination with reinforcement learning, our model can be used to discover control policies that outperform\r\nbaseline controllers. Furthermore, the recovered policies have a minimal sim-to-real gap. We showcase this by applying our control policy in-vivo on a single-layer, direct ink writing printer. "}],"department":[{"_id":"BeBi"}]}