The terms micro-robots and nanorobots get thrown around a lot in the science fiction to mean really, really small robots. Scifi also implies that the robots themselves are at the micro or nano scale and work in swarms, neither of which is necessarily true. 

In robotics,  micro- and nanorobots typically mean engineering operations that are taking place at the micro (10^-6m) or nano (10^-9 m) scale.  Engineering operations at the nano scale is essentially acting at the cellular or atomic scale, usually for 

• medical reasons (yes, the whole Fantastic Voyage idea without shrinking people, which is the point of the “Nanonauts! In Battle with Tiny Death-subs!” story by Irish writer Ian McDonald), 

• to make better chips or stronger materials (see Level 5 by William Ledbetter), or 

• both (the motivation in Prey by Michael Crichton).  

Nanobots often refer to biological protein chains using chemical processes to work on other biological protein chains  The science in Ian McDonald’s scifi story is the protein type of nanobot, but the protagonist is explaining the action using mechanical Newton-scale physics and robotics metaphors because people, even nanonauts, don’t think or relate to biological and chemical processes.

But nanorobots also mean a large device that is generating and controlling electromagnetic forces to do things like orient carbon nanotubes to produce stronger materials or

it can mean that a microelectromechanisms systems (MEMS) device, i.e., a micro-robot, is being used. MEMS aren’t as small as proteins, they are at the micro not the nano scale, but can carry sensors, like the little smart pills that a patient can swallow and be moved by magnetics when it gets near the point of interest, or can reach and work in small spaces, basically robotized extraordinarily thin wires. 

So what makes micro/nanorobots hard (besides the miniaturization and manufacturing)? Well, the change in scale changes the rules. The forces and physical properties that regular robots use to successfully vacuum a floor or weld a car aren’t the important ones. In actuation- just moving about— is hard, especially in the human body, because surface properties and adhesion dominate while gravity is less of a concern. Sensing is another challenge. Micro/nanobots generally don’t have sensors for navigation, inside the sensors are basically to act as a tiny microscope for the operator. In the case of smart pills, the designer also has to work about the wireless connection. Autonomy is not a particularly useful concept. Reproducing proteins at a site in the body is in some sense autonomous but really a mindless, automatic biological process. The same is true with the use of micro-robots for building chips. 

A great reference is:.