Doctoral Dissertations

Date of Award

12-2022

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Computer Science

Major Professor

Lynne E Parker

Committee Members

Lynne E Parker, Hairong Qi, Daniel C Rucker, Amir Sadovnik

Abstract

This research focuses on improving human-robot co-navigation for teams of robots and humans navigating together as a unit while accomplishing a desired task. Frequently, the team’s co-navigation is strongly influenced by a predefined Standard Operating Procedure (SOP), which acts as a high-level guide for where agents should go and what they should do. In this work, I introduce the concept of Constrained Collective Movement (CCM) of a team to describe how members of the team perform inter-team and intra-team navigation to execute a joint task while balancing environmental and application-specific constraints. This work advances robots’ abilities to participate along side humans in applications such as urban search and rescue, firefighters searching for people in a burning building, and military teams performing a building clearing operation. Incorporating robots on such teams could reduce the number of human lives put in danger while increasing the team’s ability to conduct beneficial tasks such as carrying life saving equipment to stranded people.

Most previous work on generating more complex collaborative navigation for human- robot teams focuses solely on using model-based methods. These methods usually suffer from the need for hard coding the rules to follow, which can require much time and domain knowledge and can lead to unnatural behavior.

This dissertation investigates merging high-level model-based knowledge representation with low-level behavior cloning to achieve CCM of a human-robot team performing collaborative co-navigation. To evaluate the approach, experiments are performed in simulation with the detail-rich game design engine Unity. Experiments show that the designed approach can learn elements of high-level behaviors with accuracies up to 88%. Additionally, the approach is shown to learn low-level robot control behaviors with accuracies up to 89%.

To the best of my knowledge, this is the first attempt to blend classical AI methods with state-of-the-art machine learning methods for human-robot team collaborative co-navigation. This not only allows for better human-robot team co-navigation, but also has implications for improving other teamwork based human-robot applications such as joint manufacturing and social assistive robotics.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS