Robust sensor-based coverage of unstructured environments

Sensor-based coverage uses sensor information to determine a path that passes a detector or some effectors over all points in an unknown space. This work identifies features of a provably complete coverage algorithm to reject "bad" sensor readings in unstructured environments without performing complicated sensor-data processing. First, we briefly review our provably complete sensor-based coverage algorithm that uses an exact cellular decomposition in terms of critical points of Morse functions. Then we present features of the algorithm that are used to overcome failures due to bad sensor data. We verified our approach by performing experiments using a mobile robot that has 16 ultrasonic sensors.