Research Areas
Below, we've broken down our research efforts into categories. There is much collaboration and overlap among these individual specialties, and many of our projects span multiple categories.
Grasping/Manipulation
What distinguishes a robot from a computing device? A crucial characteristic is its ability to interact with, and directly effect change on its environment. This interaction can take multiple forms, such as acquiring objects (grasping), placing or assembling objects, opening doors, calling an elevator, and much more. Our research aims to enable robots to interact with unstructured environments, starting from basic tasks, and building towards complex applications and true dexterity. For more, click here.
What distinguishes a robot from a computing device? A crucial characteristic is its ability to interact with, and directly effect change on its environment. This interaction can take multiple forms, such as acquiring objects (grasping), placing or assembling objects, opening doors, calling an elevator, and much more. Our research aims to enable robots to interact with unstructured environments, starting from basic tasks, and building towards complex applications and true dexterity. For more, click here.
Human-Robot Interaction
Our research approach to Human-Robot Interaction (HRI) mainly comprises quantitative experimental methods to identify important differences that impact HRI (e.g., differences between people, dimensions of robots, and various situational contexts). We opt for these methods because it is informative (and often surprising) to see what people actually do in-the-moment of interacting with a robot. By triangulating these experiments with other methods of inquiry (e.g., field studies, surveys), we aim to better address the research questions at hand. For more, click here.
Our research approach to Human-Robot Interaction (HRI) mainly comprises quantitative experimental methods to identify important differences that impact HRI (e.g., differences between people, dimensions of robots, and various situational contexts). We opt for these methods because it is informative (and often surprising) to see what people actually do in-the-moment of interacting with a robot. By triangulating these experiments with other methods of inquiry (e.g., field studies, surveys), we aim to better address the research questions at hand. For more, click here.
Motion Planning
To function effectively in noisy, real world environments, a robot must
be able to plan and execute collision-free paths in the presence of
people and clutter. Our research in motion planning is focused on
developing planners, controllers and perception modules for robust
manipulation and navigation in cluttered, dynamically-changing
environments. For more, click here.
Robot Perception
The quality of a robot's performance is critically dependent on
robust, accurate, and timely perception. Our research concentrates on
the integration of cues from cameras and laser range-finders into a
coherent world model, from geometric primitives defining the basic
structure of the world, to semantic labeling of simple and complex
objects. For more, click here.
Task Planning
Given primitives for perceiving the world and manipulating objects, a
robot must choose and sequence these basic operations to achieve its
goals. Our research focuses on task-level planning, its integration
with motion planning and perception, and robust execution. For more, click here.
