We developed an autonomous robot system that gives people step-by-step instruction on physical tasks, such as assembling a set of pipes, different instructional and communication strategies. We explored two instructional strategies: grouping and summarizing. Grouping involved the robot presenting instructions in sets of steps as opposed to individually, while summarizing involved giving a high-level summary of the goals of the steps. Additionally, the robot used mechanisms to autonomously repair any mistakes and misunderstandings. For instance, when people attached a part of the assembly incorrectly or hesitated to take action following an instruction, the robot offered clarification. The results of an evaluation of these strategies showed that grouping improved how fast the robot’s users completed the task, but it increased task breakdowns. Summarizing reduced these breakdowns. Grouping also increased the rapport between the robot and its users. We presented this work at the 2014 Robotics: Science and Systems (RSS) Conference and in a follow-up article in the Journal Autonomous Robots.
SAUPPÉ, A. & MUTLU, B. (2015). Effective Task Training Strategies for Human and Robot Instructors. Autonomous Robots, 1–17. doi: 10.1007/s10514-015-9461-0
SAUPPÉ, A. & MUTLU, B. (2014). Effective Task Training Strategies for Instructional Robots. In Proceedings of Robotics: Science and Systems (RSS 2014).