Robot motor skills can be acquired by deep reinforcement learning as neural networks to refect state–action mapping. The selection of states has been demonstrated to be crucial for successful robot motor learning. However, because of the complexity of neural networks, human insights and engineering eforts are often required to select appropriate states through qualitative approaches, such as ablation studies, without a quantitative analysis of the state importance. Here we present a systematic saliency analysis that quantitatively evaluates the relative importance of diferent feedback states for motor skills learned through deep reinforcement learning. Our approach provides a guideline to identify the most essential feedback states for robot motor learning. By using only the important states including joint positions, gravity vector and base linear and angular velocities, we demonstrate that a simulated quadruped robot can learn various robust locomotion skills. We fnd that locomotion skills learned only with important states can achieve task performance comparable to the performance of those with more states. This work provides quantitative insights into the impacts of state observations on specifc types of motor skills, enabling the learning of a wide range of motor skills with minimal sensing dependencies.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Identifying important sensory feedback for learning locomotion skills


    Beteiligte:

    Erscheinungsdatum :

    01.08.2023


    Anmerkungen:

    Nature Machine Intelligence , 5 (8) pp. 919-932. (2023)


    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch


    Klassifikation :

    DDC:    629



    CPGs with Sensory Feedback for Adaptive Bipedal Locomotion

    Habib, M.K. / Liu, G.L. / Watanabe, K. et al. | British Library Conference Proceedings | 2007


    Developing locomotion skills with deep reinforcement learning

    Peng, Xue Bin | BASE | 2017

    Freier Zugriff

    Multi-sensory tactile feedback control for robot physical interaction and biped locomotion

    Guadarrama Olvera, Julio Rogelio | TIBKAT | 2021

    Freier Zugriff

    Accessibility-Based Clustering for Efficient Learning of Locomotion Skills

    Zhang, C / Yu, W / Li, Z | BASE | 2022

    Freier Zugriff

    Advanced Skills by Learning Locomotion and Local Navigation End-to-End

    Rudin, Nikita / Hoeller, David / Bjelonic, Marko et al. | ArXiv | 2022

    Freier Zugriff