Sensor Action Planning driven by Uncertainty : Application to Object Location with Robust Local Sensors in a Nuclear Environment

This paper presents an integrated Bayesian solution to the problem of object location estimation, object recognition and sensor action planning under uncertainty. The emphasis is on finding the best next sensing action.

The method uses elementary notions from Bayesian decision theory. The best action is found as the one that optimises the expected value of a utility function, which is the logarithm of the volume of the uncertainty ellipsoid around the estimate of a target position.

An example shows that this method is capable of controlling the sensing actions of an ultrasonic sensor mounted on a robot, where the target is to accurately position a drill on a cylinder before drilling a hole.

The presented algorithm is easy to apply and computationally tractable.