My academic career started with a Bachelor of Science at the Technical University of Darmstadt. Since 2019, I am researching at the University of Cambridge where I study the joint space of immersive technologies and artificial intelligence. As of October 2022, I am a PhD student at the University of Cambridge’s Institute for Manufacturing (Department of Engineering), and part of and funded under the Engineering and Physical Sciences Research Council’s AgriFoRwArdS program1. In my PhD research, I study Digital Twins for Manufacturing Operations Management.
Research
purpose
I strive to modernise manufacturing techniques for today’s resource-limited economy using digital twins. Specifically, I build virtual environments for manufacturing operations so that employees better understand the economic consequences of activities and actions. That means I approach plugging employees into their company’s economic bigger picture. By empowering people to understand the value chain of – for example – assembly lines, I want to enable workforces to leverage their full potential. To realise this, I apply knowledge from the joint space of human interaction, digital twins, and machine learning.
Distinct
Approach
My research at the Institute for Manufacturing is unique in its focus on real cases and people. I co-founded a Learning Environment Optimisation R&D group that developed a platform to optimise Human-Computer Interaction experiments using Crowdsourcing. This platform enables my research group to experiment on a large scale with how employees can be empowered to build more knowledge, excel performance-wise, and discover new opportunities to improve manufacturing operations.
Publications
2024
Zuercher, Paul-David; Bohné, Thomas
CIRP Conference on Manufacturing Systems, vol. 57, CIRP – Collège International pour la Recherche en Productique 2024.
@conference{Zuercher2024DEPS,
title = {Discrete Event-Probabilistic Simulation (DEPS) integrated into a Reinforcement Learning Framework for Optimal Production},
author = {Paul-David Zuercher and Thomas Bohné},
url = {https://pauldavidzuercher.com/wp-content/uploads/2024/07/Discrete_Event_Probabilistic_Simulation__DEPS__integrated_into_a_Reinforcement_Learning_Framework_for_Optimal_Production-_CMS24.pdf},
year = {2024},
date = {2024-05-27},
booktitle = {CIRP Conference on Manufacturing Systems},
volume = {57},
issue = {1},
organization = {CIRP - Collège International pour la Recherche en Productique},
abstract = {This paper introduces an innovative simulation method that significantly accelerates the training of reinforcement learning agents in manufacturing, surpassing the constraints of traditional simulations. While prior Discrete Event Simulations (DES) of production lines have a runtime complexity of $O(NMS)$ for N (batched) products, M machines, and S samples, our gls{deps} can reduce an expected value computation complexity of $O(KM)$ for K time windows and M machines. Furthermore, unlike traditional DES, our developed DEPS is probabilistic enabling trustworthy decision-making by providing creditable uncertainty and risk bounds. By conditioning the product quality on machine parameters, DEPS enables rapid, risk-free AI agent training, paving the way for their integration into production lines. We (1) demonstrate that our simulation offers a state-of-the-art simulation speed, (2) provide an adaptable open-source framework for probabilistic production-line-level simulation and modelling, and (3) share a physically plausible benchmark environment. Our contribution seeks to provide a new standard for industrial gls{ai} applications, allowing manufacturers and researchers to leverage gls{rl} agents' potential for optimising production-line-level efficiency, process optimisation, and resource allocation.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
This paper introduces an innovative simulation method that significantly accelerates the training of reinforcement learning agents in manufacturing, surpassing the constraints of traditional simulations. While prior Discrete Event Simulations (DES) of production lines have a runtime complexity of $O(NMS)$ for N (batched) products, M machines, and S samples, our gls{deps} can reduce an expected value computation complexity of $O(KM)$ for K time windows and M machines. Furthermore, unlike traditional DES, our developed DEPS is probabilistic enabling trustworthy decision-making by providing creditable uncertainty and risk bounds. By conditioning the product quality on machine parameters, DEPS enables rapid, risk-free AI agent training, paving the way for their integration into production lines. We (1) demonstrate that our simulation offers a state-of-the-art simulation speed, (2) provide an adaptable open-source framework for probabilistic production-line-level simulation and modelling, and (3) share a physically plausible benchmark environment. Our contribution seeks to provide a new standard for industrial gls{ai} applications, allowing manufacturers and researchers to leverage gls{rl} agents’ potential for optimising production-line-level efficiency, process optimisation, and resource allocation.
2023
Zuercher, Paul-David Joshua; Bohné, Thomas; Hanheide, Marc
Augmenting Strawberry Agronomy: From Systematic Literature Review to Value Flow Map Working paper
2023.
@workingpaper{Zuercher2023Augmenting,
title = {Augmenting Strawberry Agronomy: From Systematic Literature Review to Value Flow Map},
author = {Paul-David Joshua Zuercher and Thomas Bohné and Marc Hanheide},
url = {https://pauldavidzuercher.com/wp-content/uploads/2023/05/Zuercher-et-al.-Augmenting-Strawberry-Agronomy-1.pdf},
year = {2023},
date = {2023-12-31},
abstract = {As the world faces the challenges of climate change and growing population, the role of agriculture in ensuring food safety is becoming more important. This paper explores how augmentation and autonomous systems can improve the efficiency of crop cultivation and support agronomists in managing the process of strawberry production. The paper first synthesises current approaches for automation and augmentation with a systematic literature review. It then identifies production and management processes and their key performance indicators. Finally, following the grounded theory approach, the paper analyses the production processes into functional key performance indicators and syntheses an agronomic value flow map of strawberry production. The comprehensive review supports researchers and agronomists to identify critical bottlenecks and offers insights into the potential for automation and augmentation in strawberry agronomy.},
keywords = {},
pubstate = {published},
tppubtype = {workingpaper}
}
As the world faces the challenges of climate change and growing population, the role of agriculture in ensuring food safety is becoming more important. This paper explores how augmentation and autonomous systems can improve the efficiency of crop cultivation and support agronomists in managing the process of strawberry production. The paper first synthesises current approaches for automation and augmentation with a systematic literature review. It then identifies production and management processes and their key performance indicators. Finally, following the grounded theory approach, the paper analyses the production processes into functional key performance indicators and syntheses an agronomic value flow map of strawberry production. The comprehensive review supports researchers and agronomists to identify critical bottlenecks and offers insights into the potential for automation and augmentation in strawberry agronomy.
Pietschmann, Leon; Zuercher, Paul-David Joshua; Bubik, Erik; Chen, Zhutian; Pfister, Hanspeter; Bohné, Thomas
Quantifying the Impact of XR Visual Guidance on User Performance Using a Large-Scale Virtual Assembly Experiment Journal Article
In: IEEE transactions on visualization and computer graphics, 2023, ISSN: 1941-0506.
@article{Pietschmann2023,
title = {Quantifying the Impact of XR Visual Guidance on User Performance Using a Large-Scale Virtual Assembly Experiment},
author = {Leon Pietschmann and Paul-David Joshua Zuercher and Erik Bubik and Zhutian Chen and Hanspeter Pfister and Thomas Bohné},
url = {https://arxiv.org/abs/2308.03390},
doi = {10.48550/arXiv.2308.03390},
issn = {1941-0506},
year = {2023},
date = {2023-08-07},
urldate = {2023-08-07},
journal = {IEEE transactions on visualization and computer graphics},
abstract = {The combination of Visual Guidance and Extended Reality (XR) technology holds the potential to greatly improve the performance of human workforces in numerous areas, particularly industrial environments. Focusing on virtual assembly tasks and making use of different forms of supportive visualisations, this study investigates the potential of XR Visual Guidance. Set in a web-based immersive environment, our results draw from a heterogeneous pool of 199 participants. This research is designed to significantly differ from previous exploratory studies, which yielded conflicting results on user performance and associated human factors. Our results clearly show the advantages of XR Visual Guidance based on an over 50% reduction in task completion times and mistakes made; this may further be enhanced and refined using specific frameworks and other forms of visualisations/Visual Guidance. Discussing the role of other factors, such as cognitive load, motivation, and usability, this paper also seeks to provide concrete avenues for future research and practical takeaways for practitioners.},
howpublished = {ArXiv},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The combination of Visual Guidance and Extended Reality (XR) technology holds the potential to greatly improve the performance of human workforces in numerous areas, particularly industrial environments. Focusing on virtual assembly tasks and making use of different forms of supportive visualisations, this study investigates the potential of XR Visual Guidance. Set in a web-based immersive environment, our results draw from a heterogeneous pool of 199 participants. This research is designed to significantly differ from previous exploratory studies, which yielded conflicting results on user performance and associated human factors. Our results clearly show the advantages of XR Visual Guidance based on an over 50% reduction in task completion times and mistakes made; this may further be enhanced and refined using specific frameworks and other forms of visualisations/Visual Guidance. Discussing the role of other factors, such as cognitive load, motivation, and usability, this paper also seeks to provide concrete avenues for future research and practical takeaways for practitioners.
Zuercher, Paul-David Joshua
Evaluating hardware differences for crowdsourcing and traditional recruiting methods Technical Report
2023.
@techreport{ZuercherEvaluatingHardwareDifferences2023,
title = {Evaluating hardware differences for crowdsourcing and traditional recruiting methods},
author = {Paul-David Joshua Zuercher},
url = {http://arxiv.org/abs/2306.09913},
doi = {10.48550/arXiv.2306.09913},
year = {2023},
date = {2023-06-16},
urldate = {2023-06-16},
abstract = {The most frequently used method to collect research data online is crowdsouring and its use continues to grow rapidly. This report investigates for the first time whether researchers also have to expect significantly different hardware performance when deploying to Amazon Mechanical Turk (MTurk). This is assessed by collecting basic hardware parameters (Operating System, GPU, and used browser) from Amazon Mechanical Turk (MTurk) and a traditional recruitment method (i.e., snowballing). The significant hardware differences between crowdsourcing participants (MTurk) and snowball recruiting are reported including relevant descriptive statistics for assessing hardware performance of 3D web applications. The report suggests that hardware differences need to be considered to obtain valid results if the designed experiment application requires graphical intense computations and relies on a coherent user experience of MTurk and more established recruitment strategies (i.e. snowballing).},
howpublished = {ArXiv},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
The most frequently used method to collect research data online is crowdsouring and its use continues to grow rapidly. This report investigates for the first time whether researchers also have to expect significantly different hardware performance when deploying to Amazon Mechanical Turk (MTurk). This is assessed by collecting basic hardware parameters (Operating System, GPU, and used browser) from Amazon Mechanical Turk (MTurk) and a traditional recruitment method (i.e., snowballing). The significant hardware differences between crowdsourcing participants (MTurk) and snowball recruiting are reported including relevant descriptive statistics for assessing hardware performance of 3D web applications. The report suggests that hardware differences need to be considered to obtain valid results if the designed experiment application requires graphical intense computations and relies on a coherent user experience of MTurk and more established recruitment strategies (i.e. snowballing).
Farr, Alexander; Pietschmann, Leon; Zürcher, Paul; Bohné, Thomas; Yapici, Guney Guven
Skill retention after desktop and head-mounted-display virtual reality training Journal Article
In: Experimental Results, vol. 4, no. 2, 2023.
@article{farr2023skill,
title = {Skill retention after desktop and head-mounted-display virtual reality training},
author = {Alexander Farr and Leon Pietschmann and Paul Zürcher and Thomas Bohné and Guney Guven Yapici},
url = {https://doi.org/10.1017/exp.2022.28},
doi = {10.1017/exp.2022.28},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Experimental Results},
volume = {4},
number = {2},
publisher = {Cambridge University Press},
abstract = {Virtual reality (VR) is increasingly used in learning and can be experienced with a head-mounted display as a 3D immersive version (immersive virtual reality [IVR]) or with a PC (or another computer) as a 2D desktop-based version (desktop virtual reality [DVR]). A research gap is the effect of IVR and DVR on learners’ skill retention. To address this gap, we designed an experiment in which learners were trained and tested for the assembly of a procedural industrial task. We found nonsignificant differences in the number of errors, the time to completion, satisfaction, self-efficacy, and motivation. The results support the view that DVR and IVR are similarly useful for learning retention. These insights may help researchers and practitioners to decide which form of VR they should use.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Virtual reality (VR) is increasingly used in learning and can be experienced with a head-mounted display as a 3D immersive version (immersive virtual reality [IVR]) or with a PC (or another computer) as a 2D desktop-based version (desktop virtual reality [DVR]). A research gap is the effect of IVR and DVR on learners’ skill retention. To address this gap, we designed an experiment in which learners were trained and tested for the assembly of a procedural industrial task. We found nonsignificant differences in the number of errors, the time to completion, satisfaction, self-efficacy, and motivation. The results support the view that DVR and IVR are similarly useful for learning retention. These insights may help researchers and practitioners to decide which form of VR they should use.
2022
Zuercher, Paul-David; Bohné, Thomas; Eger, Vera Maria; Mueller, Felix
Optimising virtual reality training in industry using crowdsourcing Conference
2022.
@conference{zuercher2022optimising,
title = {Optimising virtual reality training in industry using crowdsourcing},
author = {Paul-David Zuercher and Thomas Bohné and Vera Maria Eger and Felix Mueller},
url = {https://ssrn.com/abstract=4075130},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {Optimising virtual reality training in industry using crowdsourcing},
abstract = {The ability of Immersive Virtual Reality (IVR) to induce any training scenario in a safe and scalable manner makes it a particularly interesting technology for virtual learning factories. However, both an opportunity and a challenge is to empirically test and optimise virtual environments. Conducting scientifically robust in-person experiments is often not feasible using traditional approaches, given limited resources of training providers and near limitless opportunities to design virtual training environments. Distributed crowdsourcing techniques using Desktop Virtual Reality (DVR) with a PC may offer an alternative and more scalable approach to experimentally test and optimise virtual environments. An interesting question is therefore if such approaches using DVR are a suitable alternative to current experimental designs to enable large-scale optimisation in contexts such as virtual learning factories. While crowdsourcing has been validated for its suitability in several research applications and domains, there is limited research available on training and, to the best of our knowledge, no previous research has evaluated the suitability of crowdsourcing to optimise immersive training in industrial or learning factory contexts. With our paper we contribute the first experiment to address this research gap. Our hypothesis is that crowdsourcing is a suitable technique for IVR training optimisation if it yields equivalent results to traditional experimentation at every training optimisation level. To test this hypothesis we designed an industrial learning experiment to evaluate key performance and affective indicators of IVR training at three levels of optimisation. The experiment was conducted using traditional and crowdsourcing techniques. The results show that crowdsourcing can be a suitable alternative to traditional optimisation techniques depending on: (1) the desired operative mental state of the participants, (2) the investigated key performance indicators, and (3) the kind of optimisation performed. We contribute new data allowing important insights and an integrated training evaluation concept which can be applied when doing crowdsourcing studies.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
The ability of Immersive Virtual Reality (IVR) to induce any training scenario in a safe and scalable manner makes it a particularly interesting technology for virtual learning factories. However, both an opportunity and a challenge is to empirically test and optimise virtual environments. Conducting scientifically robust in-person experiments is often not feasible using traditional approaches, given limited resources of training providers and near limitless opportunities to design virtual training environments. Distributed crowdsourcing techniques using Desktop Virtual Reality (DVR) with a PC may offer an alternative and more scalable approach to experimentally test and optimise virtual environments. An interesting question is therefore if such approaches using DVR are a suitable alternative to current experimental designs to enable large-scale optimisation in contexts such as virtual learning factories. While crowdsourcing has been validated for its suitability in several research applications and domains, there is limited research available on training and, to the best of our knowledge, no previous research has evaluated the suitability of crowdsourcing to optimise immersive training in industrial or learning factory contexts. With our paper we contribute the first experiment to address this research gap. Our hypothesis is that crowdsourcing is a suitable technique for IVR training optimisation if it yields equivalent results to traditional experimentation at every training optimisation level. To test this hypothesis we designed an industrial learning experiment to evaluate key performance and affective indicators of IVR training at three levels of optimisation. The experiment was conducted using traditional and crowdsourcing techniques. The results show that crowdsourcing can be a suitable alternative to traditional optimisation techniques depending on: (1) the desired operative mental state of the participants, (2) the investigated key performance indicators, and (3) the kind of optimisation performed. We contribute new data allowing important insights and an integrated training evaluation concept which can be applied when doing crowdsourcing studies.
Bohné, Thomas; Heine, Ina; Mueller, Felix; Zuercher, Paul-David Joshua; Eger, Vera Maria
Gamification intensity in web-based virtual training environments and its effect on learning Journal Article
In: IEEE Transactions on Learning Technologies, pp. 19, 2022.
@article{bohne2022gamification,
title = {Gamification intensity in web-based virtual training environments and its effect on learning},
author = {Thomas Bohné and Ina Heine and Felix Mueller and Paul-David Joshua Zuercher and Vera Maria Eger},
url = {https://doi.org/10.1109/TLT.2022.3208936},
doi = {10.1109/TLT.2022.3208936},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {IEEE Transactions on Learning Technologies},
pages = {19},
publisher = {IEEE},
abstract = {Gamification approaches to learning use game-inspired design elements to improve learning. Given manifold design options to implement gamification in virtual environments, an important but underexplored research area is how the composition of gamification elements affects learning. To advance research in this area, we systematically identified key design elements that have shown promise in leading to positive learning results. We then conducted an experiment in which we varied gamification intensity in web-based virtual training environments for a procedural industrial task. 355 participants were divided into a baseline group without gamification, a basic, and an advanced gamification group. Analysis of participants' learning included learning outcomes (time-to-completion and number of mistakes), affective learning factors (motivation, self-efficacy, satisfaction), learning system usability, and perceived cognitive load throughout the learning process. The results did not show any statistically significant differences between the lower and higher levels of gamification intensity. Conversely, we found that participants’ computer gaming habits and technical equipment (display size and computer pointing device) significantly influenced learning.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Gamification approaches to learning use game-inspired design elements to improve learning. Given manifold design options to implement gamification in virtual environments, an important but underexplored research area is how the composition of gamification elements affects learning. To advance research in this area, we systematically identified key design elements that have shown promise in leading to positive learning results. We then conducted an experiment in which we varied gamification intensity in web-based virtual training environments for a procedural industrial task. 355 participants were divided into a baseline group without gamification, a basic, and an advanced gamification group. Analysis of participants’ learning included learning outcomes (time-to-completion and number of mistakes), affective learning factors (motivation, self-efficacy, satisfaction), learning system usability, and perceived cognitive load throughout the learning process. The results did not show any statistically significant differences between the lower and higher levels of gamification intensity. Conversely, we found that participants’ computer gaming habits and technical equipment (display size and computer pointing device) significantly influenced learning.
- Official EPSRC Reference: EP/S023917/1 [↩]