论文信息 - panda-gym: Open-source goal-conditioned environments for robotic learning

panda-gym: Open-source goal-conditioned environments for robotic learning

This paper presents panda-gym, a set of Reinforcement Learning (RL) environments for the Franka Emika Panda robot integrated with OpenAI Gym. Five tasks are included: reach, push, slide, pick & place and stack. They all follow a Multi-Goal RL framework, allowing to use goal-oriented RL algorithms. To foster open-research, we chose to use the open-source physics engine PyBullet. The implementation chosen for this package allows to define very easily new tasks or new robots. This paper also presents a baseline of results obtained with state-of-the-art model-free off-policy algorithms. panda-gym is open-source and freely available at https://github.com/qgallouedec/panda-gym.

[1] Jimmy Ba,et al. Adam: A Method for Stochastic Optimization , 2014, ICLR.

[2] Marcin Andrychowicz,et al. Multi-Goal Reinforcement Learning: Challenging Robotics Environments and Request for Research , 2018, ArXiv.

[3] Yuval Tassa,et al. Continuous control with deep reinforcement learning , 2015, ICLR.

[4] Konkoly Thege. Multi-criteria Reinforcement Learning , 1998 .

[5] Sergey Levine,et al. Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor , 2018, ICML.

[6] Boris Polyak,et al. Acceleration of stochastic approximation by averaging , 1992 .

[7] Herke van Hoof,et al. Addressing Function Approximation Error in Actor-Critic Methods , 2018, ICML.

[8] Silvio Savarese,et al. Making Sense of Vision and Touch: Self-Supervised Learning of Multimodal Representations for Contact-Rich Tasks , 2018, 2019 International Conference on Robotics and Automation (ICRA).

[9] Sriraam Natarajan,et al. Dynamic preferences in multi-criteria reinforcement learning , 2005, ICML.

[10] David Howard,et al. A Review of Physics Simulators for Robotic Applications , 2021, IEEE Access.