PROJECTS IN ROBOTICS

PROJECTS IN ROBOTICS: CS 6901-02 (grad) 3998-02 (undergrad) - FALL 2006

A number of independent projects for graduate and undergraduate students are available in the robotics and vision laboratory. Most of these projects require a working knowledge of either C, C++ or Java. Experience in graphics, user interfaces, computer vision or robotics is also very desirable. Students will be expected to work independently under the guidance of Professor Allen or a Ph.D. student. Projects involve using state-of-the art computer systems and robotic devices in the robotics and vision laboratory. Equipment in the lab includes include 2 RWI Pioneer Mobile Robots, an RWI ATRV2 mobile outdoor base, 2 Unimation PUMA 560 Robot Arms, 1 IBM 7575 Robot Arm, a Utah-MIT Robot Hand, a custom built overhead XYZ gantry robot, a Toshiba FMA manipulator, a Barrett Technology robotic hand, frame grabbers and imaging boards,8 SUN SPARCstations, Silicon Graphics Indigo 2 Extreme, and numerous PCs.

Interested students should contact Professor Allen in 619 CEPSR, 939-7093 or electronically at allen@cs.columbia.edu. Additional information is available at Robotics Lab Page

Below are some projects currently available in our laboratory. Students with interests other than those below are encouraged to discuss their ideas with Prof. Allen.

1. 3-D Database of Medieval France Bourbonnais Architecture

We are building a web-based searchable database of Medieval French Architecture. The database will allow students and researchers to view, analyze and compare 3-D models of these ancient structures. We need students to work with CAD tools to create 3D models of the ancient churches, and also experience in either web-interfaces or databases is desirable. see related website

2. Graspit! Robotic Hand Simulator

We have developed a robotic hand simulator called GRASPIT! that allows us to visualize and test robotic hand grasps for stability. We have recently begun working with the Cornell Neuromuscular Biomechanics Lab on a joint NSF ITR grant to merge our robotic simulation with human hand data from Cornell. Three projects related to GRASPIT! are:

We would like to create a deformable surface model to model human finger contacts. We can then implement and visualize these deformable models in the simulator
We would like to implement a kinematic model of the human hand and place it in the simulator based on the Cornell data. This project involves analyzing exisiting kinematic hand models and tuning them for our simulator.
We would like to update our machine learning algorithms to plan stable grasps. Related papers are Andrew T. Miller, Steffen Knoop, Henrik I. Christensen and Peter K. Allen, Automatic Grasp Planning using Shape Primitives, and R. Pelossof, A. Miller, P. Allen, T. Jebara, An SVM Learning Approach to Robotic Grasping

3. Columbia Mobile Site Scanning Robot

We are building a mobile site scanning robot that will be able to map the Columbia Campus and build 3-D models of the campus automatically. Projects include:

Integrate the onboard Sonar and Video to allow the robot to navigate through obstacles
Use a new "bag of pixels" learning algorithm to do topographic robot localization.
Implement a video tracking algorithm to have the robot follow a moving object
Add a remote sensor suite using an existing robotic arm system to the mobile robot (mechanical and electrical interfacing required)

4. In-vivo surgical imaging system

We are building an in-vivo surgical imaging system that is fully insertable into the abdomen. Projects include writing tracking and control software for the device. Projects include:

Integrate the onboard Sonar and Video to allow the robot to navigate through obstacles
Use a new "bag of pixels" learning algorithm to do topographic robot localization.
Implement a video tracking algorithm to have the robot follow a moving object
Add a remote sensor suite using an existing robotic arm system to the mobile robot (mechanical and electrical interfacing required)

5. Protein Manipulation Project

This project is part of a multi-institution, multi-year, NIH grant in Structural Genomics that created the NorthEast Structural Genomics Consortium This research is aimed at using vision to provide the compliance and robustness which precise protein manipulations require without the need for extensive analysis of the physics of grasping or a detailed knowledge of the environment. Tasks include visually isolating individual proteins in a culture, recovering 3-D depth models of isolated proteins, and using online visual servoing techniques to grasp and manipulate the proteins. The project team consists of researchers in Computer Science, Biomedical Engineering and Biology.

A short description of the project is available here and also a paper that appeared at Microrobotics for Biomanipulation Workshop, IROS 2003

Peter Allen (allen@cs.columbia.edu)