Inference of User’s Intentions in AdaptiveAutonomy

• Main Author: Maioli, Giovanni
• Format: Others
• Language: English
• Published: Örebro universitet, Institutionen för naturvetenskap och teknik 2020
• Subjects: Computer Sciences; Datavetenskap (datalogi)
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:oru:diva-86154

Human-Robot Interaction (HRI) is a leading topic in the field of robotics, itsimportance is manifest in those tasks where just teleoperating a robot canbe not effective, due to time delay, or lack of contest provided by the robot’ssensors. On the other hand, a pre-programmed robot can be not efficient aswell, since the environment could be unknown, and the robot could lack theability of decision making.For this reason, the aim of this project is to provide an efficient method forhuman-robot cooperation, in particular through the inference of the user’sintention while teleoperating a robotic arm.The human agent interacts with the robot through a VR-based interface,which enables him/her with a better understanding of the surrounding environment.The interface provides an interaction proxy, allowing the user to drivethe robot in the virtual reality, that reproduces the environment where therobotic arm actually is.The algorithm here implemented is able to infer the intention of the humanagent, through a method named Head-Hand Coordination Based Inference.The positions and orientations of both the head of the user and the interactionproxy are used in order to adjust the aim of the teleoperation, asexplained in detail in Chapter 3.Another result of this thesis is the capacity to detect the lack of attentionof the user, and to evaluate his/her reliability through a real value parametercalled confidence parameter.The system is responsive to the trustworthiness in the user, and is ableto slow down the robot if the human agent is distracted, as well as to movefaster if the confidence is high.In order to test the efficiency of the system, an experiment was conducted,that has shown that this method is safer than basic teleoperation, when thehuman is distracted.The data collected also reported that the operator prefers teleoperation tothe method here implemented. That comes with no surprises, since the aimiiiof the system is to reduce the control of the user when the behavior is notsafe.In the future it would be interesting to develop a similar system fromthe robotic agent side, and test the responsiveness of the human agent in asimilar HRI environment.