Repairing plans for object finding in 3-D environments

In this paper, we address the problem of repairing previously computed plans for searching for an object. The object is sought with a 7 degrees of freedom mobile manipulator robot with an ??eye-in-hand?? sensor. The sensor is limited in both range and field of view. Our method computes a set of sensing configurations, which collectively cover the environment with the 3-D visibility region of the limited sensor. An order for visiting sensing configurations, which diminishes the expected value of the time for finding the object is generated. The search plan corresponds mainly to the set of sensing configurations to be visited and the order for visiting those configurations. In this paper, we show that whenever the environment changes locally our plan can also be repaired locally. We base our approach on a 3-D convex regions decomposition dividing the environment. The plan is repaired by generating a new sub-set of sensing configurations and a new order for visiting those configurations, only considering the convex regions related to the change in the map of the 3-D environment. We have implemented all our algorithms, and we present simulation results in realistic environments.