Picking up and moving things with nano-machines is sort of the "guiding light" for many researchers. Right now, billions of dollars are going towards this effort, with many different approaches. The biggest challenge for researchers is trying to put together a LEGO structure while wearing mittens. There are tons of theoretical designs for nanobots that you can find easily on the internet, but most of these are probably not "robotic" in the sense you may be thinking of.
Right now molecular robots are more like super-synthesized molecules that just happen to react with its environment where you want it to. For example, researchers working with drug delivery will "tag" a target cell or site with a nanobot. The bot deploys a tag on a site specific protein when it comes in contact with the specific environment, like within cancer cell. You may assume the complications with this.
So, as far as molecular "robots" with autonomy or the ability to do what we want them to when we want them to... that may be a while. We will have to wait until quantum computing is more understood. You can only shrink something so much before you start to lose its original function.