Advanced robotic arm control using Kinect

[Ryan Lloyd], [Sandeep Dhull], and [Ruben D'Sa] wrote in to share a robotics project they have been keeping busy with lately. The three University of Minnesota students are using a Kinect sensor to remotely control a robotic arm, but it’s not as simple as it sounds.
Using OpenNI alongside PrimeSense, the team started out by doing some simple skeleton tracking before working with their robotic arm. The arm has five degrees of freedom, making the task of controlling it a bit tricky. The robot has quite a few joints to play with, so the trio not only tracks shoulder, elbow, and wrist movements, but they also monitor the status of the user’s hand to actuate the robot’s gripper.
When everything was said and done, the results were pretty impressive as you can see in the video below, but the team definitely sees room for improvement. Using inverse kinematics, they plan on filtering out some of the joint tracking inaccuracies that occur when the shoulders are moved in a certain way. They also plan on using a robotic arm with even more degrees of freedom to see just how well their software can perform.
Be sure to check out their site to see more details and videos.