Oliver Brock, Ph.D.
Research Associate and Scientific Consultant


Robotics Laboratory
Computer Science Department
Stanford University
Gates Building 1A, Room 122
353 Serra Mall #146
Stanford, CA 94305-9010
fon (650) 725-8810
fax (650) 725-1449
Oliver Brock

My new web page is at the University of Massachusetts Amherst.

What's on this page?

Short Bio   Research Interests   Introduction to my Research   Recent Results   Publications   Miscellaneous


Short Bio

Currently I am working with my former advisor Oussama Khatib on various aspects of elastic planning for robots with many degrees of freedom, such as humanoid robots.

Previously, I was a lecturer and research scientist at Rice University with Professor Lydia Kavraki in the Physical Computing Group. Spring 2000 I taught COMP430: Introduction to Databases; Fall 2000 I was supposed to teach COMP584: Computational Geometry, but unfortunately I had to leave Rice to fulfill other commitments back in California.

I also was a Ph.D. student at the Robotics Lab in the Computer Science department at Stanford University; my advisor was Professor Oussama Khatib.

I started out at the Technical University Berlin studying Computer Science. I wrote my master thesis in robotics in the group of Professor Günter Hommel.

back to the top


Research Interests

Currently, mobile robots and mobile manipulators often perform poorly in unknown, unstructured, and dynamic environments. By integrating motion planning and motion execution I attempt to design motion generation algorithms that combine the best of both worlds: they are immune to local minima and guaranteed to reach the goal if possible, like planning algorithms, and achieve reactive real-time obstacle avoidance, like most execution paradigms. The results of this work are the global dynamic window approach and the elastic strip framework (see publications below).

I also have been working with Lydia Kavraki on decomposition-based planning, which can be viewed as a continuation of my work on Elastic Strips and promises to be a very powerful motion planning and motion generation approach with many applications in robotics and beyond.

Other research interests include motion planning and its applications to areas like computational molecular biology, character animation, virtual prototyping; also collision avoidance, mobile manipulation, mobile robotics, and computational geometry.

back to the top


Introduction to my Research

back to the top


Recent Results

Task-Consistent Obstacle Avoidance and Motion Behavior

The elastic strip approach is an approach to real-time replanning and motion execution in dynamic environments for robots with many degrees of freedom. It allows the integration of various motion behaviors. For example, the end effector can execute a task, while the redundant degrees of freedom are used for obstacle avoidance. Below are three videos, which resulted from work together with Sriram Viji: on the left the elastic strip is used to avoid obstacles, but the task is not maintained. The video in the middle shows how task execution and obstacle avoidance can be integrated. The task is to follow the red line with the end effector. Watch how in the first video the end effector deviates significantly from the task so that the obstacles can be avoided. In the second video the end effector does not deviate at all, despite the fact that the obstacles perform the same motion as in the first video.

Sometimes constraints can make it impossible to perform the task. In the case of the third video, the end effector has to deviate from the task to avoid the second Scout robot. The framework allows task suspension and automatic task resumption; watch closely, how the end effector deviates from the line and after passing the obstacle resumes the task.

Click on the images to see the movies.

(3MB) (3MB) (3MB)

You can also check out our video submission to ICRA 2002: Task-Consistent Obstacle Avoidance for Mobile Manipulation (23MB)

(23MB)


Real-Time Obstacle Avoidance and Posture Control for Humanoid Robots

In these two videos you see the elastic strip approach applied to a humanoid robot with more than 30 degrees of freedom. This is a fairly complicated environment, in which one of the office chairs moves into the path of the robot and the upper beam of the door is lowered during the motion. The left video shows how the robot moves without posture control. The video on the right shows the same motion when a very simple posture behavior is imposed. It does not look perfectly human yet, but it serves well to illustrate the idea and the ease with which such posture can be integrated in the elastic strip framework. In the future, we will work on postures based on balance constraints and so on...

Click on the images to see the movies.

without posture (1MB) with posture (1MB)

without posture (6MB) with posture (6MB)

And here is another (slightly humorous) video, showing a humanoid robot skiing. The serious point here is to show how all degrees of freedom are used in real time to avoid obstacles. Watch how the snowman moves into the path and how the finish sign is sliding down the poles. Note how the ski poles move to avoid the snowman and to fit through the finish gate; note also how the whole posture of the robot changes to pass under the finish sign.

Click on the image to see the movie.

(5MB)

Looking at this video you can imagine how these methods find direct applicability in character animation for movies or video games. The motion of characters can be specified independently of other characters or objects in the scene, and independently of the number of degrees of freedom at the task level. This allows for more realistic games with higher interactivity for smaller production costs.

Local Replanning

And I finally got around to creating a video answering the most frequently asked question: What happens when an obstacle crosses the path and continues to move? How does the elastic strip appraoch deal with that?

(3MB)

back to the top


Publications

back to the top


Video Proceedings

back to the top


Theses

back to the top


Miscellaneous

back to the top

My new page at UMass Amherst.


  Stanford           CS Department           Robotics Lab           Manipulation Lab


Last modified on February 1, 2002 by Oliver Brock
Visitors since November 8, 2001: provided by
Elastic Strips Page
Global Dynamic Window Page
Decomposition-based Planning Page