Enabling long-term human-robot interaction: On the need for behavior coordination systems

Abstract

The deployment of service robots in private homes has become a promising area of research in robotics, bringing significant changes to how household tasks are managed and enhancing residents’ quality of life. However, as these robots become more common, a critical challenge arises in achieving effective behavior coordination to ensure smooth and safe interactions with the people in the household. To achieve long-term acceptance and integration into household routines, personalized interaction, adaptability, and continuous learning capabilities are essential. Previous research has explored two distinct approaches to deploy service robots in private homes. The first approach involves limiting the robot’s tasks to short interactions with individuals, where only a limited set of functionalities are accessible. In contrast, the second approach focuses on creating a knowledge-base that stores enough information for the behavior coordination system to make decisions on tasks based on external input. In the first approach, the robot’s capabilities are well-suited for short but relatively complex tasks within a well-prepared environment, yet it lacks the ability to explore long-term interactions without additional effort. Conversely,the second approach presents a contrasting obstacle, involving extended setup due to comprehensive architecture construction for tasks’ modeling. Nevertheless, once established, this approach enables the robot to facilitate intricate and adaptable behaviors. Indeed, the challenges discussed have contributed to the current situation in which mobile service robots are not yet widely deployed in the homes of the general public. Instead, their applications are primarily found in well-controlled environments such as museums, hotels, or care facilities, where autonomous service robots have shown success and gained acceptance. One goal is to overcome the complexities of behavior coordination and enhance their capabilities to meet the diverse and dynamic demands of household environments.To tackle these challenges, we developed an architectural middle-ground for planning, monitoring, and scheduling a set of straight forward tasks. To assess the effectiveness of our approach, we integrated the proposed architecture into five robots and conducted field trials, involving 18 users for a three-week each. Our work showcases the successful implementation of a behavior coordination system, effectively controlling mobile service robots for extended periods within private homes. Importantly, our approach goes beyond the limitations of isolated and controlled task execution found in previous research. However, we acknowledge that the system’s adaptability to individual users is still limited, despite the extensive efforts of various research groups and private enterprises searching for a comprehensive and universally applicable solution. This challenge remains, and further advancements and innovations are required to enhance the adaptability of service robots to meet the diverse needs and preferences of individual users in real-world settings.