Mems Microrobot applications evolve in ultra-dense contexts. They are at the stage of the simulation to obtain 2-Dimension or shapes for the deployment of individual or collective intelligent programs, and collectively achieve miniaturized and flexible workshops. Now-on, manufacturers and academicians projects are at the stage of modeling and prototyping Mems, and simulating Mems systems applications. To tackle these objectives, addressing individual positioning of Mems inside very dense micro-systems becomes strategic. In many situations, it is more useful to know the relative positioning between Mems and their orientation, than the knowledge of absolute positioning of each Mems. Works achieved on microrobots micropositioning are either stochastic or deterministic. The formers are based on probabilistic approaches, that provides better results on a small scale but produce greater error accumulation with a large sample. The latters are based on geometric considerations to accurately compute the position of each Mems, and then distribute it to all the elements. We propose a model of a smart-grid (orthogonal and hexagonal lattice) of microrobots with regular geometry, and their connectors (actuators and sensors for moves and other actions) communicating by contact. Based on this model, we propose a mixed positioning algorithm (absolute and relative) in 2D without mobility of Mems in a group ranging in size from thousands to millions of items, based only on neighborly relations. Then by simulation we perform a functional validation of our algorithm, and a validation of the scalability of our algorithm on orthogonal grid of over 1 million node.