Robot Learning Lab University of Calgary
Papers, datasets & preprints

Publications

Selected publications. For the complete and current list, see Google Scholar. Lab members are shown in bold.

Google Scholar ORCID Datasets

Robot learning

REBOOT Figure 1: successful trajectories and induced failures paired with expert recoveries across five assembly phases
Fig. 1 · Success and recovery, phase by phase
FeaturedDataset & benchmark · 2026

REBOOT: From Failure to Recovery – A Dataset and Benchmark for Precision Assembly

N. Y. O. Ofori-Ampofo, S. Ebrahimi Kahou, J. Thekinen
arXiv:2609.22591 [cs.RO], 2026
2,160 real-robot demonstrations across 18 precision-assembly tasks, half of them induced failures paired with expert recoveries, with per-phase benchmarks of ACT, Diffusion Policy and π0-FAST.
Design as a sequential decision process: a human designer's action changes the aircraft design from state S-t to S-t+1, and a recurrent imitation-learning policy (RNN, LSTM or GRU) learns to imitate that action. Below, aircraft designs ordered by increasing design quality, from a baseline at Q = 0 to Q = 1
Conference

Teaching AI to design from humans: a comparison of behavioral cloning architectures

G. Bozorgmehry Boozarjomehry, J. Thekinen
ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC-CIE)
Human designers' aircraft-design decisions, recorded as sequences of design states and parameter changes, used to train imitation-learning policies, with recurrent architectures (RNN, LSTM, GRU) compared on the quality of the designs they produce.

Earlier work

Learning from human decision data, crowdsourcing, and design and manufacturing systems.

GroupAnno graphical model linking instances, true labels, annotator bias and annotator-group bias to annotated labels
Fig. 1 · GroupAnno model
Conference · 2022

Toward annotator group bias in crowdsourcing

H. Liu, J. Thekinen, S. Mollaoglu, D. Tang, J. Yang, Y. Cheng, H. Liu, J. Tang
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL), Vol. 1, pp. 1797–1806, 2022
ICED framework: feature extractor, synthetic data generation and true label inference for imbalanced crowdsourced data
Fig. 1 · ICED framework
Conference · 2022

Learning from imbalanced crowdsourced labeled data

W. Wang, J. Thekinen, X. Liu, Z. Liu, J. Tang
Proceedings of the SIAM International Conference on Data Mining (SDM), pp. 594–602, 2022
Baseline and final aircraft design geometries produced by each design session
Fig. 4 · Baseline and final aircraft designs
Journal · 2021

Information exchange patterns in digital engineering: an observational study using web-based virtual design studio

J. Thekinen, P. T. Grogan
Journal of Computing and Information Science in Engineering, 21(4):041012, 2021
Nash-equilibrium regions of the two-airline route entry game as functions of unobserved variables
Fig. 1 · Nash-equilibrium regions
Journal · 2020

Modeling airlines' route selection decisions under competition: a discrete-games-based model

J. Thekinen, K. Moolchandani, J. H. Panchal, D. A. DeLaurentis
Journal of Air Transportation, 28(1):3–14, 2020
Bipartite matching of designers' parts to 3D-printer machine owners
Fig. 1.3 · Matching designers and machine owners
Dissertation · 2018

Mechanism design for complex systems: bipartite matching of designers and manufacturers, and evolution of air transportation networks

J. Thekinen
Ph.D. dissertation, School of Mechanical Engineering, Purdue University, 2018
Matching service seekers (designers) with service providers (3D-printing manufacturers) using expected utilities over part attributes
Fig. 1 · Matching designers with 3D-printing services
Journal · 2017

Resource allocation in cloud-based design and manufacturing: a mechanism design approach

J. Thekinen, J. H. Panchal
Journal of Manufacturing Systems, 43:327–338, 2017
[Add remaining publications.]