Date on Master's Thesis/Doctoral Dissertation

8-2021

Document Type

Master's Thesis

Degree Name

M.S.

Department

Electrical and Computer Engineering

Degree Program

Electrical Engineering, MS

Committee Chair

Popa, Dan O.

Committee Co-Chair (if applicable)

Welch, Karla C.

Committee Member

Welch, Karla C.

Committee Member

McIntyre, Michael L.

Committee Member

Taghavi, Nazita

Committee Member

Das, Sumit K.

Author's Keywords

autism; robotics; hri

Abstract

Robots have recently emerged as a novel approach to treating autism spectrum disorder (ASD). A robot can be programmed to interact with children with ASD in order to reinforce positive social skills in a non-threatening environment. In prior work, robots were employed in interaction sessions with ASD children, but their sensory and learning abilities were limited, while a human therapist was heavily involved in “puppeteering” the robot. The objective of this work is to create the next-generation autism robot that includes several new interactive and decision-making capabilities that are not found in prior technology. Two of the main features that this robot would need to have is the ability to quantitatively estimate the patient’s motion performance and to correctly classify their emotions. This would allow for the potential diagnosis of autism and the ability to help autistic patients practice their skills. Therefore, in this thesis, we engineered components for a human-robot interaction system and confirmed them in experiments with the robots Baxter and Zeno, the sensors Empatica E4 and Kinect, and, finally, the open-source pose estimation software OpenPose. The Empatica E4 wristband is a wearable device that collects physiological measurements in real time from a test subject. Measurements were collected from ASD patients during human-robot interaction activities. Using this data and labels of attentiveness from a trained coder, a classifier was developed that provides a prediction of the patient’s level of engagement. The classifier outputs this prediction to a robot or supervising adult, allowing for decisions during intervention activities to keep the attention of the patient with autism. The CMU Perceptual Computing Lab’s OpenPose software package enables body, face, and hand tracking using an RGB camera (e.g., web camera) or an RGB-D camera (e.g., Microsoft Kinect). Integrating OpenPose with a robot allows the robot to collect information on user motion intent and perform motion imitation. In this work, we developed such a teleoperation interface with the Baxter robot. Finally, a novel algorithm, called Segment-based Online Dynamic Time Warping (SoDTW), and metric are proposed to help in the diagnosis of ASD. Social Robot Zeno, a childlike robot developed by Hanson Robotics, was used to test this algorithm and metric. Using the proposed algorithm, it is possible to classify a subject’s motion into different speeds or to use the resulting SoDTW score to evaluate the subject’s abilities.

Download

Included in

Robotics Commons

Share

COinS