This work is on recon guration of hexagonal lattice-based modular robotic systems. The solutions are proposed in a general framework that does not exploit speci c characteristics of any particular robotic system and can be realized by many of the currently and potentially existent prototypes. In our framework, a robotic con guration is a connected set of modules which are located in a hexagonal lattice. Modules are assumed to have a simple processor and some small memory, and to be able to send and receive short messages to and from neighboring modules. Both computation and memory are assumed to be of (small) constant size, but for a module that unavoidably needs to use linear memory to store the information of the goal shape. Modules have the ability of attaching and detaching from lattice neighbors, and to rotate on attached modules. This allows them to move relative to each other and, in particular, to walk along the boundary of the robot. Within this framework, our algorithms are completely distributed, local, and synchronous. They consist in sets of rules, each one having a priority, a precondition, and an actuation or postcondition. Rules are identical for all modules, and are simultaneously executed by all them. In fact, modules are assumed to be homogeneous and indistinguishable, and the rules are run without the help of id's and without the need of any central controller or unique clock.