HTTP/2 200
content-type: text/html; charset=utf-8
x-frame-options: DENY
vary: Sec-Fetch-Dest, Sec-Fetch-Mode, Sec-Fetch-Site
cache-control: no-cache, no-store, max-age=0, must-revalidate
pragma: no-cache
expires: Mon, 01 Jan 1990 00:00:00 GMT
date: Wed, 31 Dec 2025 23:02:34 GMT
p3p: CP="This is not a P3P policy! See g.co/p3phelp for more info."
content-security-policy: base-uri 'self';object-src 'none';report-uri /_/view/cspreport;script-src 'report-sample' 'nonce-K-XAeTG_NS202kie5KSGbw' 'unsafe-inline' 'unsafe-eval';worker-src 'self';frame-ancestors https://google-admin.corp.google.com/
cross-origin-opener-policy: unsafe-none
reporting-endpoints: default="/web-reports?jobset=prod&wcrumsspbp=false&bl=editors.sites-viewer-frontend_20251210.02_p1&app=25&clss=1&context=eJwNz3tUkwUABfDPb993NQSRATpIYX5TG76alInChDXZSI18pFvaERQn4AMMEHyUwvFt1jFLM01RRCURM7GirJOJlZKmVGYUPlCc8tzGQxBNun_8zrn3_ne9LvXJ72UTSnrbhKkBNuF18ttuEwbR0902oc8nNkEqsAkBZDxlE2y3bEIylTfahD9obpBdSKO2gXZBCrELywfbhQ2UFGsXMimHNtH44deFp4QR14UQir7TIlrJkNsuRpArr13soJWLO8Q8uhjVKVbSd8ZO8Tw9dHSKT-no1k7xc5pf8UhcRG-ou8R5lCZ0i9nUOaBbFAZ2i20rusUn9MIMlSqSovNUKitNv6dSzaYVrvFSLom7oyQv-uZ4lPQjxUUYpankmWSSHtG_q1-W7lI6zFIO7VxmlvZRUrZZWkziSbPkReO-Mksm6hgeK3VTxKJYKYaGvBUrjaT1zlhpG9nHWKREuvilRaqij963SoVURGV0v8MquahgcZx0jO6qX5EaqLrnJKmWcmbky2upJiVfrqO_v86Xb1Nl6wG5irSvFch6mld9SE6ld1yH5A10JqVQLqdjSwrlU1Rxt1D-nTRfHJa1dDn-iHyNfDYclQMpKLRIHkT6N4pkA02vPybPptW0jj6LL5a_oJoFxXIdfV9ULP9ElWdL5CrKSDohr6L6pSfkFtpTUi8X0IULDfJVskc3yok051yjnETX7E3yDfLd2yT3J9U9b_Smk74-KKOhZT4YRdfP98Etyg73xRp6Nd8XM8l5uC-aqfiBH0oppd4Py6inQQ1f6qH44xkqHeuPMxQT4Y84Kp3HTuE3AjCestWBWEO94wPhTztWBWIP3d8YCBetLeiHTfRcdT88T15L-kNNXWn94ZXeH1u0GnxAyWM0qHNo4KHTJRp8R50XNRAqNHj7YBDW058FQaimnK3BWEtdrmCI7mDs6wjGYard_iwaaadpAPbQnEkDkEjXikNw4Z8QXCa9KwQGmukJwWz6ak4oflgYivNUfzoULTS5pxYzqHSslt-1qEnVorlMi4e0-tAgrKOJrkGYQkueV5BFL1kVRFLgcgUD6fEGBaqNCs5sVlBOiw8oyKTvLys4R96_KfCjDnpClisKJlMRlZDfVQX96L2nCnZSwWgdjtF_UTrIRh1ORutQRo2pOrSRe6UO7TT7Tx3mU8Lng5FMbacH4wlNsg3FdLprH4oGavHRo4ue8dOjDykj9BhGvxTocYX-cupxk8p8w3CWav3C0EhV_mGooccBYVAFhqFSiMSvxZH4g4rjo1BK3YIRPXsYEdnbiMm0bJcRK2iOZgKSqHrHBNSSQxWNReQ1JBpqqh4Rg1o6a47BVdKdj8FwWmo2YTlV7TZhV4UJ-6nuqgkeyqs3YQutCzLjXfrAYsYn1M9hhoY8S8x4RC7VRHRT4NSJGEg-bzLTzeaJcJJ_kgUaSs61IJ10hRYMp_1HLThC4y5YYKKmLCvaSTluxTBqKbGii3JvW7GZPhwVh71UfisOv9LNrXfgpIQHd7CA1NNqEUxej2qhpo9HOvEpZeQ6sYq055zQkz3tARJpaeEDLCeDdx0iaOXDOuTRvhv1OEyVjfWoIsesBpxa0YBv6f57DUgKbUIK7XI0YT81RTWjnQZ_3QzvzS4cP-nCaXrL6IYy141hlJXuxvF1btzb5sb07W58XO1G4203DC96kLjQgxRq3enBY9r4pQe76NaUFlxa1AL_Ey3Q0MyAVjgTWhE_vxU_xLRh1qo2VK9uw7TyNuTQwbkdGJPcgZ-9OxGjdELt3Wvr9b8q0Leopmpjj2BpVqojRxnhWJCalZ6ROSozNcuROTKbmyNj5MKM9LQsR9qChHBD-JjR4aMNowzhCctG_w9GO734&build-label=editors.sites-viewer-frontend_20251210.02_p1&imp-sid=CPHk-sX36JEDFRPV3gId_982Rg&is-cached-offline=false"
document-policy: include-js-call-stacks-in-crash-reports
referrer-policy: strict-origin-when-cross-origin
content-encoding: gzip
server: ESF
x-xss-protection: 0
x-content-type-options: nosniff
set-cookie: NID=527=xDVENmqpauE9o1xyNugDN30HhI2NURw1X6mV5K7Lmuu4_WRMeqN_HLB6dRD7zWyiAGfqHVaGxXTl62YFakYVdntCtavQZGilyNadPsa1mfHKB-ayPCF5gB5HCsaxXQitSR4MDL83OL6k9rfLUmkkYEZEzomq6VqUxGIR7261AjKfW4SSyrsiI5GgfB0Yy-upsl599iK6vw-Pd9Q; expires=Thu, 02-Jul-2026 23:02:34 GMT; path=/; domain=.google.com; Secure; HttpOnly; SameSite=none
alt-svc: h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
Learning Tool Morphology for Contact-Rich Manipulation Tasks with Differentiable Simulation
Learning Tool Morphology for Contact-Rich Manipulation Tasks with Differentiable Simulation
Mengxi Li, Rika Antonova, Dorsa Sadigh, Jeannette Bohg
Abstract: When humans perform contact-rich manipulation tasks, customized tools are often necessary and play an important role in simplifying the task. For instance, in our daily life, we use various utensils for handling food, such as knives, forks and spoons. Similarly, customized tools for robots may enable them to more easily perform a variety of tasks. Here, we present an end-to-end framework to automatically learn tool morphology for contact-rich manipulation tasks by leveraging differentiable physics simulators. Previous work approached this problem by introducing manually constructed priors that required detailed specification of object 3D model, grasp pose and task description to facilitate the search or optimization. In our approach, we instead only need to define the objective with respect to the task performance and enable learning a robust morphology by randomizing the task variations. The optimization is made tractable by casting this as a continual learning problem. We demonstrate the effectiveness of our method for designing new tools in several scenarios such as winding ropes, flipping a box and pushing peas onto a spoon in simulation. We also validate that the shapes discovered by our method help real robots succeed in these scenarios.
@inproceedings{ li202 3 learning ,
title={{ Learning Tool Morphology for Contact-Rich Manipulation Tasks with Differentiable Simulation} },
author={ Li, Mengxi and Antonova, Rika and Sadigh, Dorsa and Bohg, Jeannette },
booktitle={IEEE International Conference on Robotics and Automation (ICRA)},
year={2023},
organization={IEEE}
}
