Survivor search with autonomous UGVs using multimodal overt attention

Abstract

Rescue robotics is gaining increasing attention as an application of human assistance machines. Robots (or unmanned ground vehicles) for searching victims in disaster sites require a high level of reliability and robustness. This paper proposes a mobile vision system able to search for victims using integration of video and thermal camera inputs. A time efficient scanning strategy is designed for the pan-tilt camera head with a biologically inspired robust search mechanism to improve the performance of the overall system. Results show success of the proposed methodology in enhancing robustness in autonomous survivor detection.