Chengshu Li
Papers
10
Total Citations
456
H-Index
8
About
Chengshu Li is a leading researcher in embodied AI and robot learning, whose work centers on creating the simulation infrastructure and learning algorithms that enable robots to perform complex, everyday household tasks. He is the principal architect of the iGibson simulation environments (iGibson 1.0 and 2.0, with over 190 combined citations), which have become foundational tools for the field by providing large-scale, realistic, and fully interactive home scenes populated with rigid and articulated objects. These environments directly address the critical gap between simulated and real-world robotics. Li’s major contributions extend to algorithm design, most notably through ReLMoGen (over 120 citations), which innovatively integrates motion generation with reinforcement learning to solve challenging mobile manipulation tasks by lifting the action space to high-level subgoals. He has also shaped the evaluation of socially-aware robots, co-authoring key principles and guidelines for social robot navigation (over 70 citations). His recent work includes the massive BEHAVIOR-1K benchmark, a human-centered, 1,000-activity benchmark that grounds robotic tasks in real human desires. Through these efforts, Li has provided the community with the tools, tasks, and evaluation standards necessary to advance toward truly capable household robots.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4Principles and Guidelines for Evaluating Social Robot Navigation Algorithms59 citations · 2024
- 5
- 6
- 7Principles and Guidelines for Evaluating Social Robot Navigation Algorithms15 citations · 2023
- 8
- 9
- 10