Carnegie Mellon University Personal Robotics Lab

HERB

Our custom hardware platform.

HERB 2.0 HERB 2.0 is our latest bimanual mobile manipulation platform, with two Barrett WAM arms atop a completely redesigned chassis with a Segway RMP 200 base. Siddhartha Srinivasa
HERB 1.0 HERB 1.0 was our first integrated mobile manipulation platform, with the arm mounted atop the Segway. Siddhartha Srinivasa
Busboy The Busboy was our first platform for autonomous manipulation, comprised of a Barrett WAM arm bolted to the ground coordinating autonomously with a Segway RMP 200 wandering around serving drinks. Siddhartha Srinivasa

Manipulation

We focus on using physics in the design of actions, algorithms, and hands for manipulation.

Constrained Manipulation Planning Algorithms for constructing manipulator trajectories in the presence of multiple heterogeneous constraints. Dmitry Berenson
Pre-grasp Manipulation Incorporating pre-grasp manipulation such as object rotation or sliding improves the success rate and quality of transport tasks.
Best Manipulation Paper Award Finalist ICRA 2010
Lillian Chang
Caging Manipulation Constrained manipulation with caging grasps significantly increases the range of possible motions of a robot, enabling it to achieve previously infeasible tasks. Rosen Diankov
Push-grasping in Clutter A new manipulation primitive based on the mechanics of quasi-static pushing, an efficient algorithm for planning in clutter with the new primitive, and extensive demonstrations on HERB.
Best Conference Paper Award Finalist IROS 2010
Mehmet Dogar
Trajectory Optimization for Manipulation Utilizing multiple grasp targets in manipulation problems by extending functional gradient trajectory optimization methods to goal sets. Anca Dragan
Imitation Learning for Grasping Predict good grasps of novel complex free-form objects by generalizing human demonstrations. Nathan Ratliff
Grasping in Cluttered Environments Compute grasps for complex objects in cluttered scenes. Dmitry Berenson

Human-Robot Interaction

We focus on formalizing Human-Robot Interaction principles using machine learning, motion planning, and function gradient algorithms. We have been working on enabling seamless and fluent human-robot handovers.

Human-Robot Handoffs Learning spatial and temporal cues from human observation for efficient handoffs. Maya Cakmak
Gracefully Mitigating Breakdowns Expectancy-setting and recovery strategies for robots to mitigate breakdown in services.
Best Conference Paper Award Winner HRI 2010
Min Kyung Lee
People Helping Robots Helping People Crowdsourcing for grasping novel objects Alex Sorokin

Perception

Algorithms for turning sensor data into semantic information for manipulation.

MOPED A state of the art algorithm for object recognition and registration from a single image
Best Vision Paper Award Finalist ICRA 2009
Alvaro Collet
Proprioceptive Localization Using collision information opportunistically to improve manipulator and robot localization. Mehmet Dogar
GATMO An algorithm and a software package for tracking people and other movable objects. Garratt Gallagher

Joint Projects

Collaborative projects with other labs.

People Prediction Machine learning techniques for model and compute long-term and short-term trajectories of people. Drew Bagnell
Simple but General Hands Exploring the fundamental capabilities of simple hand and designing simple hands, along with control and planning techniques, to enable general purpose manipulation. Matt Mason
Model-based Hierarchical Planning Navigating at high fidelity over long ranges. Ross Knepper
Theatrical Performance Performing as an actor with a human counterpart. School of Drama, LTI, and ETC