Modeling and Improving Teleoperation Performance of Semi-Autonomous Wheeled Robots

Abstract

Robotics and unmanned vehicles have allowed us to interact with environments in ways that were impossible decades ago. As perception, decision making, and control improve, it becomes possible to automate more parts of robot operation. However, humans will remain a critical part of robot control based on preference, ethical, and technical reasons. An ongoing question will be when and how to pair humans and automation to create semi-autonomous systems. The answer to this question depends on numerous factors such as the robot's task, platform, environment conditions, and the user. The work in this dissertation focuses on modeling the impact of these factors on performance and developing improved semi-autonomous control schemes, so that robot systems can be better designed. Experiments and analysis focus on wheeled robots, however the approach taken and many of the trends could be applied to a variety of platforms.

Wheeled robots are often teleoperated over wireless communication networks. While this arrangement may be convenient, it introduces many challenges including time-varying delays and poor perception of the robot's environment that can lead to the robot colliding with objects or rolling over. With regards to semi-autonomous control, rollover prevention and obstacle avoidance behaviors are considered. In this area, two contributions are presented. The first is a rollover prevention method that uses an existing manipulator arm on-board a wheeled robot. The second is a method of approximating convex obstacle free regions for use in optimal control path planning problems.

Teleoperation conditions, including communication delays, automation, and environment layout, are considered in modeling robot operation performance. From these considerations stem three contributions. The first is a method of relating driving performance among different communication delay distributions. The second parameterizes how driving through different arrangements of obstacles relates to performance. Lastly, based on user studies, teleoperation performance is related to different conditions of communication delay, automation level, and environment arrangement. The contributions of this dissertation will assist roboticists to implement better automation and understand when to use automation.