{"title":"利用触觉反馈提高深度模仿学习模型的数据质量和数量","authors":"Catie Cuan;Allison Okamura;Mohi Khansari","doi":"10.1109/TOH.2024.3384482","DOIUrl":null,"url":null,"abstract":"Learning from demonstration is a proven technique to teach robots new skills. Data quality and quantity play a critical role in the performance of models trained using data collected from human demonstrations. In this paper we enhance an existing teleoperation data collection system with real-time haptic feedback to the human demonstrators; we observe improvements in the collected data throughput and in the performance of autonomous policies using models trained with the data. Our experimental testbed was a mobile manipulator robot that opened doors with latch handles. Evaluation of teleoperated data collection on eight real conference room doors found that adding haptic feedback improved data throughput by 6%. We additionally used the collected data to train six image-based deep imitation learning models, three with haptic feedback and three without it. These models were used to implement autonomous door-opening with the same type of robot used during data collection. A policy from a imitation learning model trained with data collected while the human demonstrators received haptic feedback performed on average 11% better than its counterpart trained with data collected without haptic feedback, indicating that haptic feedback provided during data collection resulted in improved autonomous policies.","PeriodicalId":13215,"journal":{"name":"IEEE Transactions on Haptics","volume":"17 4","pages":"984-991"},"PeriodicalIF":2.4000,"publicationDate":"2024-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Leveraging Haptic Feedback to Improve Data Quality and Quantity for Deep Imitation Learning Models\",\"authors\":\"Catie Cuan;Allison Okamura;Mohi Khansari\",\"doi\":\"10.1109/TOH.2024.3384482\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Learning from demonstration is a proven technique to teach robots new skills. Data quality and quantity play a critical role in the performance of models trained using data collected from human demonstrations. In this paper we enhance an existing teleoperation data collection system with real-time haptic feedback to the human demonstrators; we observe improvements in the collected data throughput and in the performance of autonomous policies using models trained with the data. Our experimental testbed was a mobile manipulator robot that opened doors with latch handles. Evaluation of teleoperated data collection on eight real conference room doors found that adding haptic feedback improved data throughput by 6%. We additionally used the collected data to train six image-based deep imitation learning models, three with haptic feedback and three without it. These models were used to implement autonomous door-opening with the same type of robot used during data collection. A policy from a imitation learning model trained with data collected while the human demonstrators received haptic feedback performed on average 11% better than its counterpart trained with data collected without haptic feedback, indicating that haptic feedback provided during data collection resulted in improved autonomous policies.\",\"PeriodicalId\":13215,\"journal\":{\"name\":\"IEEE Transactions on Haptics\",\"volume\":\"17 4\",\"pages\":\"984-991\"},\"PeriodicalIF\":2.4000,\"publicationDate\":\"2024-04-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE Transactions on Haptics\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://ieeexplore.ieee.org/document/10490256/\",\"RegionNum\":3,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"COMPUTER SCIENCE, CYBERNETICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Haptics","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/10490256/","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, CYBERNETICS","Score":null,"Total":0}
Leveraging Haptic Feedback to Improve Data Quality and Quantity for Deep Imitation Learning Models
Learning from demonstration is a proven technique to teach robots new skills. Data quality and quantity play a critical role in the performance of models trained using data collected from human demonstrations. In this paper we enhance an existing teleoperation data collection system with real-time haptic feedback to the human demonstrators; we observe improvements in the collected data throughput and in the performance of autonomous policies using models trained with the data. Our experimental testbed was a mobile manipulator robot that opened doors with latch handles. Evaluation of teleoperated data collection on eight real conference room doors found that adding haptic feedback improved data throughput by 6%. We additionally used the collected data to train six image-based deep imitation learning models, three with haptic feedback and three without it. These models were used to implement autonomous door-opening with the same type of robot used during data collection. A policy from a imitation learning model trained with data collected while the human demonstrators received haptic feedback performed on average 11% better than its counterpart trained with data collected without haptic feedback, indicating that haptic feedback provided during data collection resulted in improved autonomous policies.
期刊介绍:
IEEE Transactions on Haptics (ToH) is a scholarly archival journal that addresses the science, technology, and applications associated with information acquisition and object manipulation through touch. Haptic interactions relevant to this journal include all aspects of manual exploration and manipulation of objects by humans, machines and interactions between the two, performed in real, virtual, teleoperated or networked environments. Research areas of relevance to this publication include, but are not limited to, the following topics: Human haptic and multi-sensory perception and action, Aspects of motor control that explicitly pertain to human haptics, Haptic interactions via passive or active tools and machines, Devices that sense, enable, or create haptic interactions locally or at a distance, Haptic rendering and its association with graphic and auditory rendering in virtual reality, Algorithms, controls, and dynamics of haptic devices, users, and interactions between the two, Human-machine performance and safety with haptic feedback, Haptics in the context of human-computer interactions, Systems and networks using haptic devices and interactions, including multi-modal feedback, Application of the above, for example in areas such as education, rehabilitation, medicine, computer-aided design, skills training, computer games, driver controls, simulation, and visualization.