As robotic systems advance, it has been issued to manipulate various types of objects in various environment scenarios using the same robotic platform. To do that, estimating poses of objects is the pose important problem. However, robust object pose estimation in various types of environment is challenging problem. In the present thesis, I suggest robust pose estimation methods for different types of objects by using a multi-modal sensor system, an integration of RGB camera and 3D scanner. More specifically, pose estimation methods for door handles, debris, and rigid body objects are suggested. The suggested methods employed minimal human interaction in multimodal data space as well as objects priors. In addition, the 2D/3D calibration method is suggested to integrate the two sensors. The results of the experiments demonstrate the robustness of the proposed methods. Moreover, as a real-world application the suggested methods is employed as the vision system of a complete full size humanoid robot. The practical demos further validate the robustness and performance of the proposed methods.