The objective of this paper is to propose an algorithm to be used the next time a PHO scare occurs. We would therefore like to study how the algorithm would perform in case an asteroid is actually on a collision course. Luckily we do not have any real example of this situation; we resort to a simulation. The entire process of discovery, initial orbit determination and successive orbital refinement has been simulated by [Bowell and Muinonen 1992] for a hypothetical asteroid 1992 KP4 which would impact the Earth in the year 2004, only 12 years after the supposed discovery.
The simulated dataset includes observation from the discovery on May 27, 1992 until March 18, 1993 (plus pre-discovery observations in 1939). We have therefore performed a risk assessment in three phases, the first including observations only until July 6, the second one including all the simulated 1992-93 observations, the third one including the 1939 precovery observations. All the simulations about the fictitious impactor 1992 KP4 include also the Moon (in case of impact, a few thousand km can make a lot of difference).