Abstract: The project’s goal was to test the viability of using reinforcement learning for the training of a robot to pick up objects from a surface automatically. We used a WidowX robotic arm in the RepLab configuration that includes a surface on which the robot can practice picking up objects. During the project we discussed three different problems: picking up a single cube from a pre-defined location set given the location index, picking up a single cube from an unknown location on the board given a snapshot of the board, and picking up multiple cubes from unknown locations given a snapshot of the board. These problems are ordered in increasing order of difficulty and allow us to decide whether reinforcement learning is the correct tool for learning a problem such as picking up objects from a surface. We decided to use Q-learning in order to solve the first problem and TD3 in order to solve the other two problems. The first problem was solved successfully and we reached an asymptotical success rate of 100% in picking up the cube. The second problem was solved with the assistance of a supervised learning technique in order to get the cube’s location from the picture, and after that we used TD3 in order to reach that location and successfully pick up the cube. The third problem remains an open problem vis-à-vis the project but we created a wide variety of tools that allow for further research on an effective solution of the problem. In addition, we also built tools that allow convenient work on the robot like a wrapper for its source code that allows a convenient interaction with the robot, as well as simulators that create images of the robot’s surface with cubes such that the model’s training could be done on a simulation and not on the robot itself.