Micro robotic agents are more difficult to control than 'macro' robots as they suffer from a number of inherit limitations. In this work we apply a previous result on decentralized navigation to the case of micro robots. The micro-robots we are interested in, move like unicycles with bounded turning radius. To control these micro-robots, we use a multi-level control strategy. In the upper level the agents are thought as omnidirectional. The middle layer maps these movements into the motion of a unicycle, and finally the lower level drive controls the micro-agent. We present simulations to demonstrate the validity of this approach.