What happens to your autonomous or semi-autonomous robot when GPS data is not available? This has been a glaring difficulty for developers particularly of military robots.
A paper has recently been published that points to a possible way forward. Two researchers at Carnegie Mellon's Robotics Institute are proposing the use of a cost-based algorithm with kalman and particle filters to produce position estimates. The algorithm has been demonstrated using a Crusher robot with some success. The paper is presented here (PDF).
Sources: As cited
