Path-creation method to search for persons using a flying robot

Abstract

This paper proposes a path-creation method for flying robots to search for persons. The method is based on an appearance-based identification algorithm applied to the images obtained from the sky. A flying robot is a holonomic mobile; it can move in all directions, and the number of paths is infinite. The person identification was performed effectively by choosing the path suitable for the identification from within those combinations. The proposed method can create an efficient path to identify multiple persons accurately. All of the candidate paths that the flying robot can fly were estimated, then the evaluation scores of candidate paths were calculated by cumulating the identification results of multiple persons. The candidate path with the maximum evaluation score was selected. The effectiveness of the method was confirmed via simulation. The path for 1-3 persons who were within a zone of 50 m × 50 m was obtained. The time for one path segment was five seconds. Simulation of multiple path creations for the same persons was performed. Obtaining a result when there was a prior simulation was confirmed. In the simulation, the flying robot flew at a constant height of 3 m and moved at a constant speed of 30 km/h. During movement, the orientation of the camera mounted on the robot was assumed to be fixed. The camera was assumed to be a generic camera with VGA and 1 frame per second. One path creation required approximately one minute (Intel Core i7 CPU).