Volume 21 Issue 12 - Publication Date: 1 December 2002
Elastic Strips: A Framework for Motion Generation in Human Environments
Oliver Brock Laboratory for Perceptual Robotics, Computer Science Department, University of Massachusetts Amherst, Massachusetts 01003 and Ossama Khatib Robotics Laboratory, Department of Computer Science, Stanford University, Stanford, CA 94305, USA
Robotic applications are expanding into dynamic, unstructured, and populated environments. Mechanisms specifically designed to address the challenges arising in these environments, such as humanoid robots, exhibit high kinematic complexity. This creates the need for new algorithmic approaches to motion generation, capable of performing task execution and real-time obstacle avoidance in highdimensional configuration spaces. The elastic strip framework presented in this paper enables the execution of a previously planned motion in a dynamic environment for robots with many degrees of freedom. To modify a motion in reaction to changes in the environment, real-time obstacle avoidance is combined with desired posture behavior. The modification of a motion can be performed in a task-consistent manner, leaving task execution unaffected by obstacle avoidance and posture behavior. The elastic strip framework also encompasses methods to suspend task behavior when its execution becomes inconsistent with other constraints imposed on the motion. Task execution is resumed automatically, once those constraints have been removed. Experiments demonstrating these capabilities on a nine-degree-of-freedom mobile manipulator and a 34-degreeof- freedom humanoid robot are presented, proving the elastic strip framework to be a powerful and versatile task-oriented approach to real-time motion generation and motion execution for robots with a large number of degrees of freedom in dynamic environments.
Multimedia Key
= Video = Data = Code = Image
Example One: Demonstration of real-time path modification using elastic strips: The trajectory of a nine degree-of-freedom mobile manipulator (left) is modified in real time as another mobile manipulator (bottom) moves into its path. The resulting motion employs all nine degrees of freedom to accomplish obstacle avoidance, as demonstrated by the arm motion. (3.6 MB)
Example Two: This video shows the experiment from extension 1 executed on a real robot. (3.8 MB)
Example Three: The trajectory of a 34 degree-of-freedom humanoid robot is modified in real time, as it moves under a lowering beam. The video shows motion modification without posture control, leading to physically infeasible motion. (2.2 MB)
Example Four: This video shows the same experiment as extension 4 with posture control. The resulting motion is not only more natural in its appearance, but also physically feasible. (2.2 MB)
Example Five: Real-time path modification can be applied to character animation. Observe how the skiing robot with 34 degrees of freedom avoids the moving snowman and the lowering finish banner. Note how the ski poles are moved to avoid obstacles and how a natural posture is maintained with posture control. (2.7 MB)
Example Six: For this experiment the task consists of following the red line with the end-effector. Obstacle avoidance is accomplished using elastic strips and causes the end-effector to deviate significantly from the task. (3.4 MB)
Example Seven: This experiment is the same as in extention 7, except that the elastic strip performs obstacle avoidance in a task-consistent manner. You can see how the end-effector continuously performes the task. The obstacles perform the exact same motion as in the previous experiment. (3.4 MB)
Example Eight: This segment shows the expermient of extension 8 executed on the real robot. The end-effector task consists of following the red beam. Obstacle avoidance is accomplished using elastic strips in a task-consisten manner. The obstacle is perceived using a laser range finder. The last segment of this video shows task-consistent obstacle avoidance for a different task: a bottle of water is placed on a tray and the task consists of keeping the tray level during obstacle avoidance. (13.7 MB)
Example Nine: When constraints render task-consistent obstacle avoidance infeasible, the task can be suspended and subsequently resumed, as demonstrated in this video. The task consists again of following a straight-line trajectory with the end-effector. The second mobile robot moves too far into the path of the mobile manipulator for the task to be maintained. It is suspended and resumed, once the obstacle is passed. (3.4 MB)
Example Ten: Purely reactive obstacle avoidance can result in hightly suboptimal paths. This video segments shows a simple method of local replanning to avoid teh suboptimalities. (4.9 MB)
BibTeX format
Return to Contents