Annihilation by a hostile space-colonizing robot race
Not a phrase that appears very often in scientific papers, but it does appear in this paper by Max Tegmark and Nick Bostrom based on a short article published in Nature in which the authors ask "How unlikely is a doomsday catastrophe?". There are many ways to destroy the human race, but getting rid of the entire planet Earth takes a bit more effort. Still, there are some ways it could be done, some of which are mentioned in this article.
The threat from asteroids has been known about for a long time and (despite the odd scare story) nothing that is an imminent threat has been found. This doesn't mean there aren't any asteroids that will wipe us out any time soon, just that we haven't found them yet. There's a lot of stuff up there and very few telescopes looking for them.
There is also climate change, a topic that is also in the media a lot just now. Like a collision with an asteroid, this would just wipe out life on the planet rather than destroying the planet itself. Some species may survive even a major climate alteration but the Earth would be very different to how we know it today. There is still debate over whether mankind is to blame for the changes we see in the climate. No doubt we are having some effect, but we know Earth's climate changes over time anyway. How much of it is due to us, and can we do anything about it, assuming we can get everyone to agree.
There are ways that humans could destroy their own planet as well. When the Relativistic Heavy Ion Collider (RHIC), a particle accelerator used to smash particles together at the Brookhaven National Laboratory, first began operations there were reports in the media that the collisions could create black holes that would destroy the Earth and possibly the entire observable universe! Back in March there were reports that they had created a black hole, but what they actually created was a singularity, mathematically similar but without the capability to accrete matter and devour the Earth.
What Max Tegmark and Nick Bostom have done is calculate an unbiased estimate for the frequency of doomsday scenarios, those involving things outside the control of the human race such as a nearby supernova, a black hole created naturally, or invasion by a race of hostile robots. You might assume that since life on this planet has lasted so long the time between such events would be large. This would be biased though, because we have to be here to make the observation, a version of the good old anthropic principle. They point out that the frequency of these events may really be quite high, and that we have just been very lucky up until now. They conclude that the probability of a catastrophic event caused by an external event is one every billion years, and that the yearly risk from man-made accelerators is about one in a thousand billion. Very reassuring.