Skip to content. Skip to navigation
CIM Menus

School of Computer Science

Synthesis for Autonomy: Connecting the Physical, the Logical and the Human

Vasumathi Raman, Postdoctoral Scholar
Computing and Mathematical Sciences California Institute of Technology

April 7, 2015 at  10:00 AM
McConnell Engineering Room 437


Robots have revolutionized manufacturing through precise, automated manipulation of objects and tools. Deploying robots as caregivers or assistants in homes and as surrogates for humans in dangerous environments would offer even greater benefits to society. However, human settings present much higher variability and uncertainty than meticulously designed factories. Encountering previously unseen objects and environments constitutes a major challenge for autonomous robot manipulators operating in everyday human settings. The novel object manipulation problem requires a robot to grasp or otherwise manipulate objects for which it has no detailed models. In homes, hospitals, and other natural settings, groups of objects often amass, creating clutter. Such clutter further complicates robotic manipulation making even known objects difficult to detect or localize. Robots must instead discover objects present in the scene.

In this talk I will present a novel robotics approach that focuses on learning to manipulate objects through experience. In the first part of my talk, I will examine a pair of methods for discovering and recognizing objects, which respectively rely on visual and tactile sensing. These methods tightly couple perception with deliberate manipulation actions, which aid the robot in improving its estimates about the world. In the second part of my talk, I will show how a robot can autonomously learn to manipulate novel objects by generalizing its knowledge from interactions with other, previously manipulated objects. Here, I will again demonstrate results for vision and touch-based sensing. The talk concludes with a discussion of future research directions, focusing on how the coordinated interplay of perception and manipulation will enable robots to better understand and interact with the world around them.


Vasu Raman is a postdoctoral scholar in the Department of Computing and Mathematical Sciences at the California Institute of Technology. Her research explores algorithmic methods for designing and controlling autonomous cyber-physical systems, guaranteeing correctness with respect to user-defined specifications. She earned a PhD in Computer Science from Cornell University and a BA in Computer Science and Mathematics from Wellesley College.