Benchmarking the Full-Order Model Optimization Based Imitation in the Humanoid Robot Reinforcement Learning Walk

被引:0
|
作者
Chaikovskaya, Ekaterina [1 ]
Minashina, Inna [1 ]
Litvinenko, Vladimir [1 ]
Davydenko, Egor [1 ]
Makarov, Dmitry [1 ,2 ]
Danik, Yulia [1 ,2 ]
Gorbachev, Roman [1 ]
机构
[1] Moscow Inst Phys & Technol, Moscow, Russia
[2] Russian Acad Sci, Fed Res Ctr Comp Sci & Control, Moscow, Russia
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
When a gait of a bipedal robot is developed using deep reinforcement learning, reference trajectories may or may not be used. Each approach has its advantages and disadvantages, and the choice of method is up to the control developer. This paper investigates the effect of reference trajectories on locomotion learning and the resulting gaits. We implemented three gaits of a full-order anthropomorphic robot model with different reward imitation ratios, provided sim-to-sim control policy transfer, and compared the gaits in terms of robustness and energy efficiency. In addition, we conducted a qualitative analysis of the gaits by interviewing people, since our task was to create an appealing and natural gait for a humanoid robot. According to the results of the experiments, the most successful approach was the one in which the average value of rewards for imitation and adherence to command velocity per episode remained balanced throughout the training. The gait obtained with this method retains naturalness (median of 3.6 according to the user study) compared to the gait trained with imitation only (median of 4.0), while remaining robust close to the gait trained without reference trajectories.
引用
收藏
页码:206 / 211
页数:6
相关论文
共 50 条
  • [21] Behavior parameters' optimization of robot soccer based on reinforcement learning
    Gu, D.L.
    Chen, W.D.
    Xi, Y.G.
    Moshi Shibie yu Rengong Zhineng/Pattern Recognition and Artificial Intelligence, 2001, 14 (02):
  • [22] Model-Based Reinforcement Learning For Robot Control
    Li, Xiang
    Shang, Weiwei
    Cong, Shuang
    2020 5TH INTERNATIONAL CONFERENCE ON ADVANCED ROBOTICS AND MECHATRONICS (ICARM 2020), 2020, : 300 - 305
  • [23] ANN based internal model approach to motor learning for humanoid robot
    Xu, Jian-Xin
    Wang, Wei
    Vadakkepat, Prahlad
    Yee, Low Wai
    2006 IEEE INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORK PROCEEDINGS, VOLS 1-10, 2006, : 4179 - +
  • [24] Curvilinear Bipedal Walk Learning in Nao Humanoid Robot using a CPG Based Policy Gradient Method
    Shahbazi, Hamed
    Jamshidi, Kemal
    Monadjemi, Amir Hasan
    MECHANICAL AND AEROSPACE ENGINEERING, PTS 1-7, 2012, 110-116 : 5161 - 5166
  • [25] Adaptive Admittance Control for Physical Human-Robot Interaction based on Imitation and Reinforcement Learning
    Guo, Mou
    Yao, Bitao
    Ji, Zhenrui
    Xu, Wenjun
    Zhou, Zude
    2023 29TH INTERNATIONAL CONFERENCE ON MECHATRONICS AND MACHINE VISION IN PRACTICE, M2VIP 2023, 2023,
  • [26] Motion Segmentation and Recognition for Imitation Learning and Influence of Bias for Learning Walking Motion of Humanoid Robot Based on Human Demonstrated Motion
    Takahashi, Yasutake
    Hatano, Hiroki
    Maida, Yosuke
    Usui, Kazuyuki
    Maeda, Yoichiro
    JOURNAL OF ADVANCED COMPUTATIONAL INTELLIGENCE AND INTELLIGENT INFORMATICS, 2015, 19 (04) : 532 - 543
  • [27] Flexible Single-Link Manipulators Control Based on a Full-Order Transfer Function Model
    Mohsenipour, Reza
    Liu, Guangjun
    IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2024, 69 (06) : 4086 - 4093
  • [28] Learning to walk with biped robot based on an improved proximal policy optimization algorithm
    Zhang, Chao
    Zhong, Peisi
    Liang, Zhongyuan
    Liu, Mei
    Wang, Xiao
    Liu, Jinming
    INTERNATIONAL CONFERENCE ON INTELLIGENT EQUIPMENT AND SPECIAL ROBOTS (ICIESR 2021), 2021, 12127
  • [29] Voltage optimization strategy for a distribution network based on deep reinforcement learning considering regionalization and imitation learning
    Li, Shidan
    Li, Hang
    Li, Guojie
    Han, Bei
    Xu, Jin
    Li, Ling
    Wang, Hongtao
    Dianli Xitong Baohu yu Kongzhi/Power System Protection and Control, 2024, 52 (22): : 1 - 11
  • [30] Research on Foothold Optimization of the Quadruped Crawling Robot based on Reinforcement Learning
    Liu X.
    Wang P.
    Dong R.
    Recent Patents on Mechanical Engineering, 2024, 17 (01) : 11 - 22