Max Robotics — Global Service Robot Directory + US Certification

About

Caiwei Song’s research lies at the intersection of robot learning, motion planning, and safe physical human-robot interaction. Her work focuses on enabling robots to acquire complex skills from demonstration and then robustly execute those skills in unstructured, dynamic environments. A key contribution is her development of a framework that combines unsupervised trajectory segmentation with movement primitives, allowing robots to learn intricate motion sequences without manual labeling. This work, her most cited paper (30 citations), provides a scalable approach to robot skill acquisition. Song has also pioneered methods for reactive task execution, addressing the critical challenge of real-time collision avoidance when a robot operates outside its training environment. Her 2018 paper on this topic proposes a control architecture that ensures safety during learned tasks. Further extending this line of inquiry, she has developed a general control framework for physical contact tasks, using invariance control and multi-priority strategies to manage the tradeoff between safety and performance under external disturbances. Her earlier work includes the design of a novel SMA-actuated earthworm-like robot, demonstrating her breadth from bio-inspired hardware to advanced control theory.