PR2: A Physics- and Photo-realistic Testbed for Embodied AI and Humanoid Robots

Hangxin Liu,Qi Xie,Zeyu Zhang,Tao Yuan,Xiaokun Leng,Lining Sun,Song-Chun Zhu,Jingwen Zhang,Zhicheng He,Yao Su

2024-09-03

Abstract:This paper presents the development of a Physics-realistic and Photo-\underline{r}ealistic humanoid robot testbed, PR2, to facilitate collaborative research between Embodied Artificial Intelligence (Embodied AI) and robotics. PR2 offers high-quality scene rendering and robot dynamic simulation, enabling (i) the creation of diverse scenes using various digital assets, (ii) the integration of advanced perception or foundation models, and (iii) the implementation of planning and control algorithms for dynamic humanoid robot behaviors based on environmental feedback. The beta version of PR2 has been deployed for the simulation track of a nationwide full-size humanoid robot competition for college students, attracting 137 teams and over 400 participants within four months. This competition covered traditional tasks in bipedal walking, as well as novel challenges in loco-manipulation and language-instruction-based object search, marking a first for public college robotics competitions. A retrospective analysis of the competition suggests that future events should emphasize the integration of locomotion with manipulation and perception. By making the PR2 testbed publicly available at <a class="link-external link-https" href="https://github.com/pr2-humanoid/PR2-Platform" rel="external noopener nofollow">this https URL</a>, we aim to further advance education and training in humanoid robotics.

Robotics

What problem does this paper attempt to address?

The paper aims to address the following issues:

1. **Lack of full-size humanoid robot simulation platforms**: Existing robot simulation platforms (such as Gazebo, Webots, etc.) support high-fidelity physical simulation but lack photorealistic rendering in the environment, which limits the integrated research of robot perception and action. Meanwhile, platforms focused on Embodied AI overly simplify the simulation of robot dynamic behavior.

2. **Insufficient accessibility of educational and research resources**: Mini humanoid robots traditionally used (such as DARwIn-OP and NAO) have limited servo drive and load capacity, making it difficult to explore advanced topics; some low-cost force/torque-controlled legged robot platforms are limited in popularity due to cost and maintenance difficulties.

3. **Customization and high cost of competition platforms**: Although high-level competitions like the DARPA Robotics Challenge have promoted the development of humanoid robot testing platforms, these platforms are usually customized for specific teams, making them unavailable to others, very expensive, and requiring high expertise.

To address these issues, the authors propose PR2, a humanoid robot testing platform that combines photorealistic rendering and physically realistic simulation, aiming to promote collaborative research between Embodied AI and robotics, and improve accessibility and cost-effectiveness in education and research. Additionally, the effectiveness of the platform was validated by hosting a national university student humanoid robot competition, which attracted a large number of participating teams. Based on the competition results, the platform was improved for public release.