Project Abstract
We design a physics-based algorithm to produce the human-like robot motion of operating multiple objects while walking in the environment. One of the inputs to our algorithm is spatial constraints which describe the features of the environment as well as properties of objects in the environment, for
instance, the start and end position, the direction to approach the object, etc. Another input is a manipulation graph which defines the end effectors that the robot can use and the transition between end effectors. The output is a robot motion walking through the environment as planned and doing
multiple tasks simultaneously. Our algorithm exploits task forces to determine when and how to do multiple tasks like maintaining, transporting, reaching while having human characteristic. The motion produced via forces are consistent with dynamics of tasks without any reference data.
Related Work
- O. Khatib, L. Sentis, J. Park, and J. Warren. "Whole body dynamic behavior and control of human-like robots". International Journal of Humanoid Robotics, pages 29–44, 2004.
- O. Khatib, L. Sentis."Prioritized Multiobjective Dynamics And Control of Robots in Human Environments". Humanoids, pages 764-780, 2004/
- O. Khatib."A Unified Approach for Motion and Force Control of Robot Manipulators:The Operational Space Formulation".IEEE Journal of Robotics and Automation, pages 43-53, 1987
- Y. Abe and J. Popovic."Interactive Animation of Dynamic Manipulation". Proceedings of the 2006 ACM SIGGRAPH/Eurographics symposium on Computer animation, 2006
Progression
(1) Khatib and Yeuhi Controllers for tracking
Khatib's
Yeuhi's
(2) The Controllers being used for some of the tasks we are studying
Transporting, with shoulder posture maintenance control
Transporting with posture maintenance
(3) The first stumbling block was eliminating this sort of motion from the maintain object position tasks
(4) Testing of true multitasking with maintenance and pick up tasks
(5) Walking visualization of the location planner algorithm
(6) Final stages of the multitask controller
Controller performing pick and maintenance and posture tasks without excessive motion
More than two end-effectors are now being used to perform simultaneous tasks of picking up, maintaining and posture.
