Demonstration Videos for NASA Dual Arm Phase I SBIR

The following videos exhibit work accomplished by TRACLabs for the NASA Phase I SBIR, "A Planning and Control Toolkit for Dual Arm Manipulation".
Click on an image thumbnail to view an entire video (m4v format).

1 MB
Demo-1
An early implementation of simulated, dual-arm planning. This simulation had no physics, so our 3 finger hands could easily pick up a box without opening. This demonstrates planning that constrains the motion of both arms in order to pickup and move the box while keeping the box's orientation stable.

3.5 MB
Demo-2
In this physics-based simulation, we removed the hands to maximize friction in order to move the box. This shows that the box may move slightly due to gravity, despite using the exact same planning and control that was used in the idealized Demo-1 scenario above.

15 MB
Demo-3
This shows the same constrained planning and control from Demo-1 and Demo-2, but the algorithms are running on our real-world dual-arm robot.

4.2 MB
Demo-4
This demonstrates the same planning and control but with an obstacle that requires a more complicated trajectory.

6.7 MB
Demo-5
This illustrates how the planner takes observations of local obstacles into an Octree in order to plan collision-free trajectories.

15 MB
Demo-6
This demo shows how adding a previously unseen obstacle to the robot's manipulation space triggers a halt to any control that was based on a plan using outdated perception data.

55 MB
Demo-7
This demonstrates vision-in-the-loop dual-arm manipulation. The box is covered with visual fiducials to inform the robot of it's pose in space. After picking up the box, the robot returns to it's origin and waits a moment before dropping the box.

118 MB
Demo-8
This extends on the work in Demo-7. After locating the box, the robot realizes that in order to grasp the box, it must first navigate through an obstacle field. It does so using it's LIDAR and SLAM methods. Here the robot is a bit conservative in its positioning and takes a while to position itself precisely in front of the box before grasping.

23 MB
Demo-9
Here the robot uses it's planar LIDAR to detect a line of unevenly spaced rails on a rolling platform. It uses both hands to move the platform to the right, keeping at least one hand on a pole at any time.


Updated: August 27, 2012