Based on our previous work, this paper proposes an improved robotic door-opening framework. The improved framework has three components, grasping pattern, door-opening pattern, and semantic monitoring and exception handling. (1) Grasping pattern contains various grasping points and ways for different types of door knob. The point cloud of the target area is captured using Kinect sensor, and the door edge and knob is then extracted. (2) Determination of door-opening pattern includes sub operations and the corresponding trajectories. (3) By using semantic monitoring and exception handling, the tactile data from the gripper during the task execution is monitored based on which exceptional situation is assessed and handled. The proposed framework was validated by two normal door-opening experiments and two exception handling experiments using KUKA prototype.