Date on Master's Thesis/Doctoral Dissertation

8-2024

Document Type

Master's Thesis

Degree Name

M.S.

Department

Electrical and Computer Engineering

Degree Program

Electrical Engineering, MS

Committee Chair

Popa, Dan

Committee Co-Chair (if applicable)

McIntyre, Micheal

Committee Member

McIntyre, Micheal

Committee Member

Chitalia, Yash

Author's Keywords

robotics; pHRI; NAC; SODTW; motion similarity; skin sensors

Abstract

An essential part of robotics research is human-robot collaboration, which enables the use of current and new robots in everyday life and the workforce. This research applies to both parts of human-robot collaboration: physical human-robot interaction (pHRI), as well as non-physical human-robot interaction. The physical interaction uses tactile sensors and a Neuroadaptive Controller (NAC) to allow for the guidance of a robotic arm and its end-effector. The non-physical interaction uses a novel motion similarity metric, the Cartesian Segment Online Dynamic Time-Warping (SODTW), to allow a robot to better adapt to the speed of the user performing the motion during imitation exercises. Physical human-robot interaction allows for a human to touch or interact with the robot in a physical manner. The research focusing on pHRI in this thesis was conducted through robot skin tactile sensors. These tactile sensors are fabricated as arrays using a piezoresistive polymer, PEDOT: PSS. 3x3 or 4x4 sensor patches fabricated for this project are named "SkinCells". Eight SkinCells were attached to a structural handle in the shape of a can, named the OctoCan. The OctoCan was connected to the end-effector of a robot forming a handle used to guide the robot arm in the direction of force applied. A Neuroadaptive controller (NAC) enhanced the interaction between the OctoCan and the user by compensating for robot and sensor nonlinearities. This controller consists of two loops. The outer loop calibrates a filter on the skin sensors to help provide a smooth trajectory and account for sensor degradation. Sensor tuning was accomplished using an Auto-Regressive Moving Average (ARMA) algorithm trained by Recursive Least-Squares. The inner loop of our NAC controller uses a neural network in succession with an adaptive PD controller to allow the robot manipulator to follow a trajectory smoothly even when the robot dynamics change. Once the outer loop had undergone a one-time calibration process, the controller was deployed to move the arm in the direction of the force applied. The research on non-physical interaction between humans and robots uses a small humanoid robot with 4-DoF in its arms. This robot, Zeno, was designed to help diagnose autism spectrum disorder by determining the motion similarity between itself and the user. Zeno performs a motion with its arms, such as a hand wave, and then the user's movement is recorded using an Xbox 360 Kinect. Once Zeno has performed a motion, Dynamic Time Warping (DTW) determines motion similarity by producing a cost. The lower the cost between the two time series, the more similar the motions are. Using DTW in both joint space and Cartesian space can give a better understanding of how well the user is matching the robot. Once one motion cycle is completed, DTW is used to compare against a slow, normal, and fast version of the motion. The lowest cost of the three speeds is then used to adjust to the user's speed of motion. For both research topics, an adaptive approach to how humans interact with robots was employed, with the ultimate goal of creating a safer and richer experience with the robot during human-robot collaboration. This is done with the OctoCan by allowing for physical guidance of an end-effector alongside a neuroadaptive controller and is achieved with Zeno by allowing for a change in the speed of a motion by using motion similarity with dynamic time warping.

Download

Included in

Robotics Commons

Share

COinS