Well learning methods are a matter of personal style, so whatever works for you. But it's important to come away with thorough understanding at the end, rather than "I tried X and it worked, so X must be the right way to do it".
Anyway your question is so vague as it stands, that I can't make concrete suggestions. How about this: what are some embedded things you have already done, and what parts interested you? Or, why do you want to program embedded in the first place?
Embedded programming is often just a resource-constrained version of programming bigger computers. Unless you're sort of forced into that by end-goal requirements, you might as well program bigger computers so you can solve bigger problems, have more convenient dev tools, etc. If you're pursuing embedded without having an end goal that needs it, it sounds to me like you're just drifting.
(Added:) Hmm ok, here is one idea that I think is cool: get some popular commercial quadcopter and replace the controller code with something that you release as FOSS. That would give more freedom to lots of quadcopter users. There is something like that on an old Adacore page, but I don't think it fits existing popular hardware.
Another: watch some "Stuff Made Here" videos on youtube for inspiration. That guy builds stuff with a lot of code inside.
Also not mentioned: what is your budget? For a pure software project you are allowed to say "zero" since all you need is time and keystrokes. But for embedded, if you do anything interesting, you will end up accreting a workbench full of oscilloscopes, SMT rework tools, and so on. That takes $$$ not just for the gear itself, but for the living space that it will occupy if you are in a high rent location. One reason I don't do this stuff is I don't have the space for it.