• Media type: E-Book; Electronic Thesis; Doctoral Thesis
  • Title: Action understanding and prediction during inter-agent interaction
  • Contributor: Cheng, Minghao [Author]
  • imprint: Georg-August-Universität Göttingen: eDiss, 2024-06-11
  • Language: English
  • DOI: https://doi.org/10.53846/goediss-10539
  • Keywords: Informatik (PPN619939052) ; Human interaction ; Human robots interaction
  • Origination:
  • Footnote: Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
  • Description: The utilisation of robotics has seen a significant rise over the years, particularly fuelled by the burgeoning advancements in deep learning and artificial intelligence research and development. Despite this progress, the challenge of social acceptance remains a significant hurdle in the widespread adoption of robots for practical applications. A prominent aspect of this challenge lies in the effectiveness and naturalness of HRI (Human-Robot Interaction). Enhancing robotic behaviours holds considerable potential to substantially improve the social acceptance of robots. The approach to achieve the enhancement of HRI of this research is by firstly obtaining a profound comprehension of human behaviours in interactive situations of HHI (Human-Human Interaction). Subsequently, the research will use the unveiled insights and principles to guide the design, evolution, and implementation of socially adept robots. The research begins by crafting experiment paradigms centred on the concept of "double contingency," which is a term stands for a social situation that the agents perceive each other without having prior knowledge about the forthcoming events. Across the study, three distinct experiment games are designed to simulate diverse interaction scenarios. Game 1 delves into competitive contexts, employing the do/undo paradigm and assigning two player roles: leader and follower. It challenges follower players to keenly observe the actions of the leader player and counteract accordingly. Game 2 and Game 3 target collaborative scenarios, where two players are tasked with constructing a predefined combination of objects while constrained to act simultaneously. Subsequently, the research targets at implementing a sophisticated experiment setup capable of capturing details of human behaviour patterns during interaction. Eye movement, touching/untouching events and hand movement are the three major types of data in this research. The successful implementations have enabled the experiment setup to effectively and sufficiently ...
  • Access State: Open Access
  • Rights information: Attribution (CC BY)