Kaijen Hsiao- Research Scientist

Dr. Kaijen Hsiao is a Research Scientist at Willow Garage, working on grasping, manipulation, and planning under uncertainty.  Kaijen's current focus is on reactive grasp adjustment and reasoning about uncertainty and geometric constraints while doing pick-and-place operations.  Her overall goal is to make robots less clumsy in unstructured environments (like your home).  This requires careful reasoning about the current state of the world and its uncertainties and constraints (Where is the object I want to pick up?  Maybe it's over here instead?  Or maybe it's not quite the shape I thought it was.  Did it shift around in my hand after I grasped it?) as well as the likely results of actions given those uncertainties and constraints (Am I likely to knock this object over because I'm not sure precisely where it is?  Can I still put it in this little cubby-hole if I pick it up this way?).  Reasoning explicitly about uncertainty and geometric constraints is especially necessary for fiddly tasks that require precision, like grasping objects by their handles, using tools, or performing assembly tasks.

Kaijen received her Ph.D. in Computer Science from MIT in 2009, under the supervision of Tomas Lozano-Perez and Leslie Kaelbling.  Her thesis, "Relatively Robust Grasping," studied how grasping can be made more robust using touch sensing, in the face of significant object pose uncertainty.  The thesis proposes the use of POMDP models for manipulation, which allow the robot to intelligently select actions that gather information until certain that the grasp will succeed with high probability.  Kaijen has also worked on using IR fingertip sensors for reactive grasping, grasp adaptation, and imitation learning of grasp trajectories.

View Kaijen Hsiao's publications