Robotics: Science and Systems XVI

Learning Memory-Based Control for Human-Scale Bipedal Locomotion

Jonah Siekmann, Srikar Valluri, Jeremy Dao, Francis Bermillo, Helei Duan, Alan Fern, Jonathan Hurst


Controlling a non-statically stable biped is a difficult problem largely due to the complex hybrid dynamics involved. Recent work has demonstrated the effectiveness of reinforcement learning (RL) for simulation-based training of neural network controllers that successfully transfer to real bipeds. The existing work, however, has primarily used simple memoryless network architectures, even though more sophisticated architectures, such as those including memory, often yield superior performance in other RL domains. In this work, we consider recurrent neural networks (RNNs) for sim-to-real biped locomotion, allowing for policies that learn to use internal memory to model important physical properties. We show that while RNNs are able to significantly outperform memoryless policies in simulation, they do not exhibit superior behavior on the real biped due to overfitting to the simulation physics unless trained using dynamics randomization to prevent overfitting; this leads to consistently better sim-to-real transfer. We also show that RNNs could use their learned memory states to perform online system identification by encoding parameters of the dynamics into memory.



    AUTHOR    = {Jonah Siekmann AND Srikar Valluri AND Jeremy Dao AND Francis Bermillo AND Helei Duan AND Alan Fern AND Jonathan Hurst}, 
    TITLE     = {{Learning Memory-Based Control for Human-Scale Bipedal Locomotion}}, 
    BOOKTITLE = {Proceedings of Robotics: Science and Systems}, 
    YEAR      = {2020}, 
    ADDRESS   = {Corvalis, Oregon, USA}, 
    MONTH     = {July}, 
    DOI       = {10.15607/RSS.2020.XVI.031}