Measurement of fish's learning speed using antagonistic relation of robot and fish

This paper describes fish's intelligent strategies for survival observed in the hostile situation between robot and fish. The robot pursues catching fish by a net attached at robot's hand through visual servoing of hand-eye manipulator, whose real-time tracking method utilizes an optimization technique of a genetic algorithm (GA) in order to track swimming fish hastily and precisely. Through catching/releasing experiments conducted consecutively, it has been shown the fact that the fish possess an ability to produce avoiding behaviors by their own intelligence against a robot that constantly tracks the fish to try to catch them by net. This means that the fish have come up with ideas to escape from the exhaustless robot while saving their limited energy. In this paper we propose a method to measure fish's learning speed by comparing the behavior of fish with that of robots taking consistent hostile action of catching fish.