Researchers at Georgia Tech (labmates of this author) have developed a robot that can robustly open closed doors. The target application for the robot, named El-E ("Ellie"), is assistive tasks related to healthcare in the homes of the disabled. This application demonstrates a set of behaviors that enable a mobile manipulator to reliably open a variety of doors and traverse doorways using force-sensing fingers and a laser range finder.

To use the robot, the operator selects the door knob with a standard green laser pointer. The robot perceives the 3D location of the handle. After determining the handle's location, the robot drives up to the door using both the laser point's 3D location and the laser scanner on the front of the robot.  After it approaches the door, the robot searches with its fingers to find the door knob. The sensors at the base of the fingers give the robot the capability to interact the door by measuring the forces and torques present on the fingers. 

   

The robot's fingers have small six-axis force-torque sensors at their base to enable estimation of touch. This perception is used to determine if the handle is unlocked, which end of the handle to twist, and the angle of the handle. Once the robot has unlatched the door, it uses feedback from the fingers' tips to ensure it is safe to open the door. The results from the experiments show that it is possible to have the robot use the same behavior to open different doors, see image below.

Video of the robot operating on a door that opens from the other side: 

The work was published at the workshop Robot Manipulation: Intelligence in Human Environments at the 2008 Robotics: Science and Systems Conference (RSS 2008). For more detail on the robot and experiments on door openning, see the paper published on this approach. Images and video courtesy of Advait Jain, Georgia Institute of Technology.