Fundamental Challenges in Deep Learning for Stiff Contact Dynamics

被引:17
|
作者
Parmar, Mihir [1 ]
Halm, Mathew [1 ]
Posa, Michael [1 ]
机构
[1] Univ Penn, Grasp Lab, Philadelphia, PA 19104 USA
基金
美国国家科学基金会;
关键词
D O I
10.1109/IROS51168.2021.9636383
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Frictional contact has been extensively studied as the core underlying behavior of legged locomotion and manipulation, and its nearly-discontinuous nature makes planning and control difficult even when an accurate model of the robot is available. Here, we present empirical evidence that learning an accurate model in the first place can be confounded by contact, as modern deep learning approaches are not designed to capture this non-smoothness. We isolate the effects of contact's non-smoothness by varying the mechanical stiffness of a compliant contact simulator. Even for a simple system, we find that stiffness alone dramatically degrades training processes, generalization, and data-efficiency. Our results raise serious questions about simulated testing environments which do not accurately reflect the stiffness of rigid robotic hardware. Significant additional investigation will be necessary to fully understand and mitigate these effects, and we suggest several avenues for future study.
引用
收藏
页码:5181 / 5188
页数:8
相关论文
共 50 条
  • [21] "Contact" of Nanoscale Stiff Films
    Yang, Fut K.
    Zhang, Wei
    Han, Yougun
    Yoffe, Serge
    Cho, Yungchi
    Zhao, Boxin
    LANGMUIR, 2012, 28 (25) : 9562 - 9572
  • [22] Understanding deep learning - challenges and prospects
    Adnan, Niha
    Umer, Fahad
    JOURNAL OF THE PAKISTAN MEDICAL ASSOCIATION, 2022, 72 (02) : S66 - S70
  • [23] Deep Learning in Cybersecurity: Challenges and Approaches
    Imamverdiyev, Yadigar N.
    Abdullayeva, Fargana J.
    INTERNATIONAL JOURNAL OF CYBER WARFARE AND TERRORISM, 2020, 10 (02) : 82 - 105
  • [24] Data Management Challenges for Deep Learning
    Raj, Aiswarya
    Bosch, Jan
    Olsson, Helena Holmstrom
    Arpteg, Anders
    Brinne, Bjorn
    2019 45TH EUROMICRO CONFERENCE ON SOFTWARE ENGINEERING AND ADVANCED APPLICATIONS (SEAA 2019), 2019, : 140 - 147
  • [25] Challenges in Deep Learning for Multimodal Applications
    Ghosh, Sayan
    ICMI'15: PROCEEDINGS OF THE 2015 ACM INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION, 2015, : 611 - 615
  • [26] Deep Learning in Neuroimaging: Promises and challenges
    Yan, Weizheng
    Qu, Gang
    Hu, Wenxing
    Abrol, Anees
    Cai, Biao
    Qiao, Chen
    Plis, Sergey M.
    Wang, Yu-Ping
    Sui, Jing
    Calhoun, Vince D.
    IEEE SIGNAL PROCESSING MAGAZINE, 2022, 39 (02) : 87 - 98
  • [27] Software Engineering Challenges of Deep Learning
    Arpteg, Anders
    Brinne, Bjorn
    Crnkovic-Friis, Luka
    Bosch, Jan
    44TH EUROMICRO CONFERENCE ON SOFTWARE ENGINEERING AND ADVANCED APPLICATIONS (SEAA 2018), 2018, : 50 - 59
  • [28] Deep Learning in Automotive: Challenges and Opportunities
    Falcini, Fabio
    Lami, Giuseppe
    SOFTWARE PROCESS IMPROVEMENT AND CAPABILITY DETERMINATION, SPICE 2017, 2017, 770 : 279 - 288
  • [29] Challenges concerning deep learning in SPOCs
    Filius, Renee M.
    de Kleijn, Renske A. M.
    Uijl, Sabine G.
    Prins, Frans J.
    van Rijen, Harold V. M.
    Grobbee, Diederick E.
    INTERNATIONAL JOURNAL OF TECHNOLOGY ENHANCED LEARNING, 2018, 10 (1-2) : 111 - 127
  • [30] Challenges for the Repeatability of Deep Learning Models
    Alahmari, Saeed S.
    Goldgof, Dmitry B.
    Mouton, Peter R.
    Hall, Lawrence O.
    IEEE ACCESS, 2020, 8 : 211860 - 211868