This thesis presents a method for opening and closing drawers with a robotic arm based on the calculation of global trajectory using computer vision and the force and position control during robot manipulation. The methods were implemented and tested on a 6DoF UR5 robotic manipulator with a Robotiq 3-finger gripper. Computer vision methods from the RVL program library are used to detect the drawer model, processing images obtained from an Intel RealSense LiDAR L515 camera attached to the robot manipulator end effector. Based on the model of the drawer and the mutual position of the parts of the robot manipulator and the drawer, the trajectory for opening and closing the drawer is calculated. Force and position control is performed by reading the values of the built-in force and torque sensor FT 300-S. To ensure robust manipulation of the drawer, the force is minimized in the lateral directions, while position control is performed in the direction of opening and closing the drawer. In this way, adaptive corrections of the local opening and closing trajectory are ensured. The proposed solution consists of three phases: 1. recognition of the drawer model based on the recorded RGB-D image sequence of a human demonstration of opening the drawer, 2. calculation of the position for grasping the handle with a robot gripper, and 3. force and position controlled robot action of opening and closing the drawer, according to the calculated trajectory. The experimental evaluation showed that the success of the action of opening and closing the drawer was 62.5%. It was found that the recognition of the model depends on the position of the camera at the time of capturing the image of the drawer on which the recognition is performed, and that the most frequent errors occurred due to incorrect recognition of the model, i.e. due to an error in the part of the program based on computer vision. The system was implemented and tested in the Ubuntu operating system, the ROS Noetic distribution and within the Docker container.