Q. Davis, Stephanie J. Woodman, Melanie Landesberg, Rebecca Kramer‐Bottiglio, J. Bongard
{"title":"Subtract to Adapt: Autotomic Robots","authors":"Q. Davis, Stephanie J. Woodman, Melanie Landesberg, Rebecca Kramer‐Bottiglio, J. Bongard","doi":"10.1109/RoboSoft55895.2023.10122102","DOIUrl":null,"url":null,"abstract":"Robot adaptation is typically limited to adaptive control policies or actuated morphology changes (such as shape change). When part of a robot body is removed it is typically viewed as an injury that must be adapted to; the potential for adaptation through subtraction by removal of body components has not yet been considered. Biological systems, on the other hand, provide many examples of subtractive adaptation, including gene or nucleotide deletion at the evolutionary scale, apoptosis at the cellular scale, and autotomy (the deliberate loss of an appendage) at the organismal scale. In this work, we consider the adaptive potential of evolved autotomy in simulated soft robots. To do so we jointly evolved the body plans, control policies, and/or which body parts to remove for soft robots. Our results show that autotomy, rather than policy adaptation, sometimes evolved to change the robot's heading when commanded. In most trials, policy adaptation was favored by the evolutionary algorithm over autotomy for changing heading. But the fact that autotomy appeared as a viable solution in some evolving populations, both when starting body plans were evolved or set manually, suggests that this form of morphological adaptation may be useful for future soft robots.","PeriodicalId":250981,"journal":{"name":"2023 IEEE International Conference on Soft Robotics (RoboSoft)","volume":"154 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 IEEE International Conference on Soft Robotics (RoboSoft)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/RoboSoft55895.2023.10122102","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Robot adaptation is typically limited to adaptive control policies or actuated morphology changes (such as shape change). When part of a robot body is removed it is typically viewed as an injury that must be adapted to; the potential for adaptation through subtraction by removal of body components has not yet been considered. Biological systems, on the other hand, provide many examples of subtractive adaptation, including gene or nucleotide deletion at the evolutionary scale, apoptosis at the cellular scale, and autotomy (the deliberate loss of an appendage) at the organismal scale. In this work, we consider the adaptive potential of evolved autotomy in simulated soft robots. To do so we jointly evolved the body plans, control policies, and/or which body parts to remove for soft robots. Our results show that autotomy, rather than policy adaptation, sometimes evolved to change the robot's heading when commanded. In most trials, policy adaptation was favored by the evolutionary algorithm over autotomy for changing heading. But the fact that autotomy appeared as a viable solution in some evolving populations, both when starting body plans were evolved or set manually, suggests that this form of morphological adaptation may be useful for future soft robots.