Optimised task allocation using dynamic production data in human-robot teams
The demand of both industrial and consumer customers for increasingly higher degrees of customisation in products will see greater amounts of high mix production in the future of manufacturing. Despite this, automation must be implemented to improve the efficiency and output of manufacturing process...
| Main Author: | |
|---|---|
| Format: | Thesis (University of Nottingham only) |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://eprints.nottingham.ac.uk/64432/ |
| _version_ | 1848800131994877952 |
|---|---|
| author | Smith, Thomas Anthony |
| author_facet | Smith, Thomas Anthony |
| author_sort | Smith, Thomas Anthony |
| building | Nottingham Research Data Repository |
| collection | Online Access |
| description | The demand of both industrial and consumer customers for increasingly higher degrees of customisation in products will see greater amounts of high mix production in the future of manufacturing. Despite this, automation must be implemented to improve the efficiency and output of manufacturing processes. However, traditional automation methods are often unsuitable due to long lead times for setup and little flexibility to adapt them to new tasks. Human-Robot (HR) teams provide a potential way to implement easily reconfigurable automation into future factories by utilising the best characteristics of human workers such as adaptability and intelligence with those of robot workers such as strength and repeatability. Robust task planning is required to implement such HR teams. However, current approaches allow adaptation to change in performance or composition of HR teams or optimisation of tasks as a whole but not necessarily both.
In this research, a novel generalised task planning framework is proposed that uses a semi-online task planning approach, utilising online production data to determine worker capabilities then planning a manufacturing task for the HR team offline between task iterations. A system architecture is defined for such a framework but the focus of this research is the development and testing of the core technologies required for the framework to function to assess its utility. These include dynamic cost functions utilising online production data to accurately quantify the capabilities of human and robot workers across a work shift. These use continuous variables to quantify gradual changes in worker performance across a work shift; and discrete variables to detect instantaneous changes in capabilities that occur during a single task iteration. Additionally, a dynamic task planner is developed that implements dual layers of the Discrete Gravitational Search Algorithm to search for an optimum set of task assignments and task plan for a HR team given worker costs. Finally, mechanisms are proposed to intelligently implement task replanning across a work shift to optimise a HR team’s performance whilst ensuring it does not occur too frequently or unnecessarily.
These core technologies were tested individually in example cases then combined together to test the ability of the task planning framework to optimise the performance of a HR team in two example manufacturing tasks across simulated work shifts. This showed that the dynamic cost functions represent an effective way to quantify and detect any changes in a worker’s capabilities across a work shift. Additionally, task replanning was shown to improve the performance of the HR team in some scenarios, such as the human worker being over fatigued, by reassigning subtasks to the robot worker as their performance declines. Importantly, the proposed task planning framework represents a generalised methodology that can easily be redeployed to different manufacturing tasks or compositions of HR teams. |
| first_indexed | 2025-11-14T20:46:42Z |
| format | Thesis (University of Nottingham only) |
| id | nottingham-64432 |
| institution | University of Nottingham Malaysia Campus |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-14T20:46:42Z |
| publishDate | 2021 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | nottingham-644322021-07-31T04:40:27Z https://eprints.nottingham.ac.uk/64432/ Optimised task allocation using dynamic production data in human-robot teams Smith, Thomas Anthony The demand of both industrial and consumer customers for increasingly higher degrees of customisation in products will see greater amounts of high mix production in the future of manufacturing. Despite this, automation must be implemented to improve the efficiency and output of manufacturing processes. However, traditional automation methods are often unsuitable due to long lead times for setup and little flexibility to adapt them to new tasks. Human-Robot (HR) teams provide a potential way to implement easily reconfigurable automation into future factories by utilising the best characteristics of human workers such as adaptability and intelligence with those of robot workers such as strength and repeatability. Robust task planning is required to implement such HR teams. However, current approaches allow adaptation to change in performance or composition of HR teams or optimisation of tasks as a whole but not necessarily both. In this research, a novel generalised task planning framework is proposed that uses a semi-online task planning approach, utilising online production data to determine worker capabilities then planning a manufacturing task for the HR team offline between task iterations. A system architecture is defined for such a framework but the focus of this research is the development and testing of the core technologies required for the framework to function to assess its utility. These include dynamic cost functions utilising online production data to accurately quantify the capabilities of human and robot workers across a work shift. These use continuous variables to quantify gradual changes in worker performance across a work shift; and discrete variables to detect instantaneous changes in capabilities that occur during a single task iteration. Additionally, a dynamic task planner is developed that implements dual layers of the Discrete Gravitational Search Algorithm to search for an optimum set of task assignments and task plan for a HR team given worker costs. Finally, mechanisms are proposed to intelligently implement task replanning across a work shift to optimise a HR team’s performance whilst ensuring it does not occur too frequently or unnecessarily. These core technologies were tested individually in example cases then combined together to test the ability of the task planning framework to optimise the performance of a HR team in two example manufacturing tasks across simulated work shifts. This showed that the dynamic cost functions represent an effective way to quantify and detect any changes in a worker’s capabilities across a work shift. Additionally, task replanning was shown to improve the performance of the HR team in some scenarios, such as the human worker being over fatigued, by reassigning subtasks to the robot worker as their performance declines. Importantly, the proposed task planning framework represents a generalised methodology that can easily be redeployed to different manufacturing tasks or compositions of HR teams. 2021-07-31 Thesis (University of Nottingham only) NonPeerReviewed application/pdf en cc_by https://eprints.nottingham.ac.uk/64432/1/Thomas%20Smith%20PhD%20Thesis%20Submission.pdf Smith, Thomas Anthony (2021) Optimised task allocation using dynamic production data in human-robot teams. PhD thesis, University of Nottingham. Human-robot interaction; Adaptive computing systems; Task analysis; Job analysis |
| spellingShingle | Human-robot interaction; Adaptive computing systems; Task analysis; Job analysis Smith, Thomas Anthony Optimised task allocation using dynamic production data in human-robot teams |
| title | Optimised task allocation using dynamic production data in human-robot teams |
| title_full | Optimised task allocation using dynamic production data in human-robot teams |
| title_fullStr | Optimised task allocation using dynamic production data in human-robot teams |
| title_full_unstemmed | Optimised task allocation using dynamic production data in human-robot teams |
| title_short | Optimised task allocation using dynamic production data in human-robot teams |
| title_sort | optimised task allocation using dynamic production data in human-robot teams |
| topic | Human-robot interaction; Adaptive computing systems; Task analysis; Job analysis |
| url | https://eprints.nottingham.ac.uk/64432/ |