First Evaluation of a Mechanism to Open Doors with the APR-01 Assistant Personal Robot

This paper proposes and evaluates a mechanism to open doors with the APR-01 assistant personal robot to enhance its versatility in indoor environments. The specific manipulator is designed as a dedicated mechanical hand to reach, hold and turn the lever of a conventional door handle. The mechanism is first evaluated under manual control to validate its operability, and then a procedure for automatic door handle detection is proposed to enable autonomous door opening. In manual control mode, a user remotely operates the mobile robot and activates the mechanical hand guided by the images provided by a camera located in the robot head. Since the results on manual control were successful, the feasibility of automatic door handle detection is then assessed. In autonomous mode, the mobile robot analyzes the close-range images captured by a camera embedded in the manipulator. This process is performed using a k-means clustering procedure, which automatically classifies relevant elements to detect and locate possible door handles in the image. Results from door handle detection were also successful, suggesting that this information could be used by the robot to self-adjust its position and control the mechanical arm to open doors autonomously.