Back to search

University of Leeds

Improving task difficulty modelling for robot teaming in multi-dimensional contexts: with applications in performance prediction and LLM-driven multi-robot planning

Abstract

dc:description.abstract

This thesis introduces a novel framework for modelling task difficulty in Human-Machine Teaming (HMT), inspired by Fitts’ Law and extended into a six-degrees-of-freedom spatial domain. The proposed method accounts for both translational and rotational constraints between machine agents and their targets, enabling precise HMT performance prediction in complex, real-world tasks. By integrating cognitive fatigue into the model using the SAFTE (Sleep, Activity, Fatigue, and Task Effectiveness) framework, the approach holistically captures both long-term skill levels and short-term cognitive effectiveness of human operators. This enables realistic and adaptive forecasting of team performance under varying operational demands. The framework supports multiple applications. First, it provides a robust predictive model for HMT performance, useful for mission planning, workload estimation, and system adaptation. Second, it enables quantitative evaluation through a hybrid scheme combining objective measures, predictive curves, and subjective assessments (e.g., NASA-TLX, SUS). The model has been validated through a comprehensive human study, encompassing a simulation and real-world experiments involving the teleoperation of a quadruped mobile manipulator with different interfaces. Finally, this task difficulty model is adapted to support decision-making in large language model (LLM)-driven multi-robot task allocation. We introduce FittsPrompt, a pre-processing scheme that abstracts spatial complexity into structured difficulty descriptors. This abstraction allows LLMs to make more efficient and scalable task allocation decisions compared to raw observation inputs. Evaluations across 42 open- and closed-source LLMs demonstrate that the proposed approach not only surpasses traditional baseline methods but also outperforms expert human planners in real-world robotic task allocation.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Leeds
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Wan, Yuhui
Advisors dc:contributor.advisor
  • Richardson, Robert
  • Barber, Andrew
  • Jackson-Mills, George
  • Zhou, Chengxu
ORCID person:identifier.orcid

Rights

Language dc:language.dcterms_iso639_2
eng

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Wan, Yuhui. Improving task difficulty modelling for robot teaming in multi-dimensional contexts: with applications in performance prediction and LLM-driven multi-robot planning. doctoral thesis, University of Leeds, 2025.