Context-Sensitive Personalities and Behaviors for Robots

Abstract

This paper proposes Context-Sensitive Behaviors for Robots (CSBR), a method for generating diverse behaviors for robots in indoor environments based on five personality traits. This method is based on a novel model developed in this work that reacts to a synthetic genome that defines the personality of the robot. The model functions return different answers and reactions, depending on a given spoken request. The responses of the robot included spoken answers, facial animations, gestures, and actions. The novelty of this method lies in its capacity to adapt the behavior of the robot according to the context of the request. Moreover, the model is scalable since its functions not only return spoken answers but also physical responses, such as opening a gripper, saying hello with gestures, or animating a face that represents an emotion according to the context. Changes in the parameters of the synthetic genome produce different behaviors. By defining different synthetic genomes, robots can adapt to different people's moods. In this work, we introduce two scenarios for human–robot interaction in two domestic environments (house and office) through spoken requests from a human user. We implemented our method in Care-O-Bot 4 and defined three synthetic genomes to produce three behaviors: friendly, detached, and hostile. In the considered scenarios, we asked the robot the same set of requests for every synthetic genome. Not only did Care-O-Bot 4 answer according to its personality, but it also proved that our method produces different behaviors. For these scenarios, we assume that the given request includes its connotation. Since our method has characteristics influenced by context, we show that the robot's behavior changed according to the human mood and the environment.