A Framework for Improving the Speed and Performance of Teleoperated Mobile Manipulators.

Abstract

Despite recent advances in robot autonomy, teleoperation remains an integral part of many robot tasks. In situations where it is hazardous or difficult for humans to be present, but which require human judgment and decision-making skills, the use of a human operator is the only option. However, there are many issues resulting from limited feedback channels that degrade perception and manipulation abilities in remote environments, causing even basic robot tasks to be difficult and time-consuming. For robots to become more useful tools for humans in remote environments, the speed and ease of teleoperated tasks must be increased.

This purpose of this dissertation is to develop a framework for increasing speed and performance of teleoperated mobile robot tasks. First, the key issues affecting teleoperated robot system performance are defined and characterized. These factors are incorporated into an optimization-based approach for evaluating multiple design options for teleoperated systems. This optimization may require models for system components that are not readily available, and must be estimated or measured empirically.

Modeling user performance in teleoperation tasks can be particularly difficult. This dissertation focuses on obtaining such models by performing several user studies designed to predict the teleoperator performance in response to multiple manual input devices and visual feedback mechanisms, as well as varying system latencies.

The overall framework for improving system performance is based on incorporating the derived, estimated, and measured component models into the implementation of the design optimization over a series of operations in the teleoperation system's required task set.

The contributions of this dissertation are as follows: 1) An identification of the factors limiting teleoperation system performance. 2) A framework for performing design optimization of teleoperated mobile robot speed and performance. 3) An evaluation of teleoperator performance with two different manual interfaces and two different visualization interfaces. 4) The development of a performance model for a path-following steering task under different latency conditions that indicates a possible mapping between performance under constant latency and variable latency. 5) The development and validation of a driver model capable of generating human-like steering inputs to a mobile robot.