A friend of mine developed a more efficient version of RANSAC and named it ANTSAC [1] which basically follows the same biological motivation (ants that follow similar paths based on pheromone information). In the case of ANTSAC it tries to optimize the next sample guesstimation.
Thanks for posting this interesting link - these kind of old-skool AI algorithms like ant colony optimisation don't crop up in HN very often nowadays :)
[1] http://www.cogsys.cs.uni-tuebingen.de/forschung/terminiert/A...