Recycling is becoming more crucial due to the fast pace of consumerism. This thesis explores how well a robot arm, with three degrees of freedom, can be implemented to give an autonomous recycling process. After the prototyping phase it was found that a cylindrical robot arm was best suited for the project. Computer vision in addition with machine learning is used for sorting and detecting objects. The end effector is a suction cup, connected to a plastic 60-milliliter syringe. Negative pressure is created by pulling and pushing a lead screw connected to a stepper motor. The accuracy of the ML-model, the robot’s movement andmax weight are evaluated. The ML-model is trained to detect four classes; plastic, metal, paper, and glass. The thesis found that the ML-model could classify plastic the most sufficient and paper the least. The robot arm’s movement had an average error of 0.54 cm and the maximum weight was 900 grams. For future development it would be interesting to compare a range of different suction cups, to see how the material, diameter, and depth would affect its ability to pick up various objects. Additionally, the model could be enhanced by training it on a larger dataset.