A methodology for operationalising the robot centric HRI paradigm : enabling robots to leverage sociocontextual cues during human-robot interaction

Publication Type:
Issue Date:
Full metadata record
The presence of social robots in society is increasing rapidly as their reach expands into more roles which are useful in our everyday lives. Many of these new roles require them to embody capabilities which were typically not accounted for in traditional Human-Robot Interaction (HRI) paradigms, for example increased agency and the ability to lead interactions and resolve ambiguity in situations of naïvety. The ability of such robots to leverage sociocontextual cues (i.e. non-verbal cues dependent on the social-interaction space and contextual-task space in order to be interpreted) is an important aspect of achieving these goals effectively and in a socially sensitive manner. This thesis presents a methodology which can be drawn on to successfully operationalise a contemporary paradigm of HRI – Kirchner & Alempijevic’s Robot Centric HRI paradigm – which frames the interaction between humans and robots as a loop, incorporating additional feedback mechanisms to enable robots to leverage sociocontextual cues. Given the complexities of human behaviour and the dynamics of interaction, this is a non-trivial task. The Robot Centric HRI paradigm and methodology were therefore developed, explored and verified through a series of real-world HRI studies (ntotal = 435 = 16 + 24 + 26 + 96 + 189 + 84). Firstly, by drawing on the methodology, it is demonstrated that sociocontextual cues can be successfully leveraged to increase the effectiveness of HRI in both directions of communication between humans and robots via the paradigm. Specifically, cues issued by social robots are shown to be recognisable to people, who generally respond to them in line with human-issued cues. Further, enabling robots to read interaction partners’ cues in situ is shown to be highly valuable to HRI, for example by enabling robots to intentionally and effectively issue cues. In light of the finding that people will display HHI-predicted sociocontextual cues such as gaze around robots, a novel head yaw estimation framework which showed promise for the HRI space was developed and evaluated. This enables robots to read human-issued gaze cues and mutual attention in situ. Next, it is illustrated that a robot’s effectiveness at achieving its goal(s) can be increased by adding to its ability to moderate the cues it issues based on information read from humans (i.e. increased interactivity). Finally, the above findings are shown to generalise to other sociocontextual cues, social robots and application spaces, demonstrating that the developed methodology can be drawn on to successfully operationalise the Robot Centric HRI paradigm, enabling robots to leverage sociocontextual cues to more effectively achieve their goal(s) and meet the requirements of their expanding roles.
Please use this identifier to cite or link to this item: