Developing a simulation for research on semi-autonomous robot control in a restaurant environment
L. Bannink (TU Delft - Industrial Design Engineering)
J.H. Boyle – Mentor (TU Delft - Industrial Design Engineering)
O. Siebinga – Mentor (TU Delft - Industrial Design Engineering)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Labour shortages in the restaurant sector have increased interest in the use of service robots to support daily operations. While autonomous delivery robots can reduce physical workload, they struggle to operate effectively in socially complex environments. Fully tele operated robots,
including systems controlled remotely by human operators, have shown limited efficiency, as operators spend a significant amount of time monitoring robots that do not require direct human input.
This creates a need for shared control systems in which multiple semi autonomous robots operate independently, while a single human operator is supported in identifying situations where human intervention adds the most value. Such systems should provide both an overview of robot
activity and the means to selectively intervene through remote control, allowing human and robotic strengths to be combined more effectively.
This thesis presents a simulated restaurant environment designed as a research platform for studying shared control of semi autonomous service robots. The simulation enables the use of multiple visualisation methods to guide operator attention, primarily based on machine vision
derived metrics, and supports a range of tele operation control strategies. The platform is designed to be modular, allowing additional visualisations, control methods, and evaluation metrics to be integrated for future research.
The secondary goal was to kickstart this research by testing a set of visualisation methods against each other in an automated testing setup. This allows automatic data collection, such as
a total human obstruction score, movement analysis, and a secondary task in the form of math questions to determine the cognitive load of every method. However, due to some unreliability in the data caused by inaccuracies in the human obstruction calculation, none of the options
showed a clear numerical advantage over the others.
Thus, the primary contribution of this work is a configurable simulation environment that enables systematic investigation of shared control strategies for service robots in socially complex spaces. The platform provides researchers with tools to prototype, test, and evaluate human robot
collaboration methods under controlled yet realistic conditions.