Learning Emergent Gaits with Decentralized Phase Oscillators - ICRA 2024
Title: Learning Emergent Gaits with Decentralized Phase Oscillators: on the role of Observations, Rewards, and Feedback
Abstract: We present a minimal phase oscillator model for learning quadrupedal locomotion. Each of the four oscillators is coupled only to itself and its corresponding leg through local feedback of the ground reaction force, which can be interpreted as an observer feedback gain. The oscillator itself is interpreted as a latent contact state-estimator. Through a systematic ablation study, we show that the combination of phase observations, simple phase-based rewards, and the local feedback dynamics induces policies that exhibit emergent gait preferences, while using a reduced set of simple rewards, and without prescribing a specific gait.
code: https://github.com/mit-biomimetics/ORCAgym
preprint: https://arxiv.org/abs/2402.08662
Видео Learning Emergent Gaits with Decentralized Phase Oscillators - ICRA 2024 канала Biomimetics MIT
Abstract: We present a minimal phase oscillator model for learning quadrupedal locomotion. Each of the four oscillators is coupled only to itself and its corresponding leg through local feedback of the ground reaction force, which can be interpreted as an observer feedback gain. The oscillator itself is interpreted as a latent contact state-estimator. Through a systematic ablation study, we show that the combination of phase observations, simple phase-based rewards, and the local feedback dynamics induces policies that exhibit emergent gait preferences, while using a reduced set of simple rewards, and without prescribing a specific gait.
code: https://github.com/mit-biomimetics/ORCAgym
preprint: https://arxiv.org/abs/2402.08662
Видео Learning Emergent Gaits with Decentralized Phase Oscillators - ICRA 2024 канала Biomimetics MIT
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Skating Robot "RoboBlade" Video](https://i.ytimg.com/vi/ESHWhNSeFmo/default.jpg)
![Spinybot 2004](https://i.ytimg.com/vi/VKxTuKSM_rE/default.jpg)
![Impulse-based controller acceleration with step down](https://i.ytimg.com/vi/i62xnCQrGac/default.jpg)
![Prototype Cheetah Footpad demonstrating Normal and Shear Force Sensing](https://i.ytimg.com/vi/_6Rvv9s5hko/default.jpg)
![Bounding gait with Dynamic CPG](https://i.ytimg.com/vi/ZrkdU7zvRK4/default.jpg)
![Retraction Jump - 20kg 2D platform with front leg](https://i.ytimg.com/vi/dMM3RVypCvw/default.jpg)
!["Hopping for Apples" Simulation](https://i.ytimg.com/vi/nGyus_jd-RM/default.jpg)
![Robot Arm "Puncher" Video](https://i.ytimg.com/vi/z1wbjIMUs8o/default.jpg)
![iSprawl - running at 15 bodylengths/sec (5 mph)](https://i.ytimg.com/vi/Vd_2V4mGzzI/default.jpg)
![MIT Cheetah robot Walking simulation](https://i.ytimg.com/vi/Gq6fl3lTlug/default.jpg)
![Hopping Robot "DeBug" Video](https://i.ytimg.com/vi/zx86lAXD-OQ/default.jpg)
![forcesensor](https://i.ytimg.com/vi/VMh1HOhe-hQ/default.jpg)
![Neck injury](https://i.ytimg.com/vi/9KU7amuNJqg/default.jpg)
![Geckochuck](https://i.ytimg.com/vi/rC4KLKkoVYU/default.jpg)
![mcgeer-like limit cycle.avi](https://i.ytimg.com/vi/90AqnMC2upg/default.jpg)
![High Force Bilateral Haptics.mp4](https://i.ytimg.com/vi/AM5wGM0883Q/default.jpg)
![K-Robot Limit Cycle](https://i.ytimg.com/vi/_N5ar4G8ALc/default.jpg)
![MIT Cheetah: New Design Paradigm Shift toward Mobile Robots](https://i.ytimg.com/vi/R0dgELUeBTU/default.jpg)
![MIT Cheetah runs at 14.4km/h by SODJW algorithm.](https://i.ytimg.com/vi/L-OXhfp3tWk/default.jpg)
![Hopping Robot "DeBug" Simulation](https://i.ytimg.com/vi/4ue7-329pfo/default.jpg)