Robotic Autonomy and Collaboration
ARMLab performs research in advancing the capabilites of autonomous robots and making them effective collaborators.
Research Topics
Collaboration and Shared Autonomy
Description:
In this work we investigate instances where a human and robot may have shared control or need to transition control of the motion of a robot or vehicle. The robot must understand the scene and potential danger, and when necessary take control from the human to maintain safety. The robot should also make its strategy legible to the human for the human to take over if necessary and hand-over control safely.
For collaboration, it is often necessary that the behavior of the teammate (human or robot) be modeled for prediction towards collaboration. This requires understanding the teammates observation capabilities, policy, and control authority in the context of the mutual task.
Papers and Videos:
Collaboration from Observation: Human-Robot Cooperative Transport
Description:
In this project, we investigated how a robot could learn to collaborate with a human by observing two humans work together in a data-driven approach. We studied this in the context of cooperatively transporting an object, generative models are able to capture human-teammate behavior conditioned on the state of the task and surrounding obstacles and can be transferred to the robot in a human-robot team.
