Programmable self-assembly has been widely studied because of its capability to create ordered patterns from a group of multiple disordered agents without an external controller. To achieve this, assembly units must exhibit different characteristics: they need to be small, to have the ability to latch and unlatch, and low-power consumption. In addition, they need to be easily programmable and able to communicate with each other. This thesis presents an experimental robotic platform for programmable self-assembly. In this work, we build in the Usbot modular robotic cubes making use of their advantages and simplicity as its passive magnetic latching mechanism, and we endow them with communication capabilities. The system allows only local communication between the modules, specifically with the most recent linked neighbor cube. The transmission of the relevant cube data is performed by a pair of LED and ambient light sensors in a binary format. The different experiments demonstrate and compare distributed programmable self-assembly using various algorithms from the literature as Singleton and Lynchpin.