Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics
Thesis (MEng)--Stellenbosch University, 2026.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2026
|
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614028529926144 |
|---|---|
| access_status_str | Open Access |
| author | Renner, Daniel Stefan |
| author2 | Louw, Louis. |
| author_browse | Louw, Louis. Renner, Daniel Stefan |
| author_facet | Louw, Louis. Renner, Daniel Stefan |
| author_sort | Renner, Daniel Stefan |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2026. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/135903 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:45:31.220Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2026 |
| publishDateRange | 2026 |
| publishDateSort | 2026 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/135903 Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics Renner, Daniel Stefan Louw, Louis. Palm, D. Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Thesis (MEng)--Stellenbosch University, 2026. Renner, D. S. 2026. Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/bbe86ac0-7db0-411a-9eeb-95ca324aff4c Virtual assistants are widely used in consumer settings but remain rare in industrial intralogistics. Existing systems often rely on predefined commands, struggle with specialised terminology, and lack flexibility in dynamic environments. At the same time, intralogistics is characterised by complex human interaction, diverse technical systems, and changing workflows. These conditions create a need for virtual assistants that can interpret natural language instructions that are not predefined, understand context, and autonomously perform multi-step tasks across different technical systems. Insights from the literature review led to the hypothesis that such capabilities may be realised through a virtual assistant based on a large language model and a system of multiple cooperating agents. The literature review further confirmed that no existing concept combines these elements in intralogistics, establishing a clear research gap. The aim of this study is to design and evaluate such a concept. The guiding research question is: How can a concept for a virtual assistant in intralogistics that processes non-predefined natural language input through a system of multiple cooperating agents based on a large language model be realised? The study follows a design science research approach. A prototype was built based on the proposed concept. The evaluation followed the framework for evaluation in design science with a summative and artificial strategy. Testing took place in a controlled learning-factory environment. The prototype processed natural language instructions covering three central task types in intralogistics: information management, collaborative activities, and knowledge transfer. It successfully interpreted all variations of these tasks, identified required actions, coordinated tool use through its agent-based structure, returned correct structured outputs, and executed the corresponding behaviour. The prototype achieved a one hundred percent success rate across all tested tasks. These findings show that a system based on a large language model with multiple cooperating agents enables a virtual assistant to operate without predefined commands and to handle the defined complexities of intralogistics. The study concludes that the proposed concept is technically feasible and fulfils the defined requirements. It provides a viable way to enable natural language interaction in intralogistics and can match the functional scope of existing solutions while improving usability and flexibility. Limitations include testing in a laboratory environment, text-only interaction without speech input, no support for user profiles, and no testing of tolerance to misspellings or noisy language. Future work should evaluate performance in real-world environments, integrate speech-based interaction, increase robustness to misspelled input, strengthen data connections, and address further evaluation dimensions such as response time, safety, and trustworthiness. Masters 2026-04-14T13:00:50Z 2026-04-14T13:00:50Z 2026-03 Thesis https://scholar.sun.ac.za/handle/10019.1/135903 en Stellenbosch University 217 pages application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Renner, Daniel Stefan Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title | Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title_full | Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title_fullStr | Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title_full_unstemmed | Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title_short | Concept for a virtual assistant with an LLMbased multi-agent system to process user input in intralogistics |
| title_sort | concept for a virtual assistant with an llmbased multi agent system to process user input in intralogistics |
| url | https://scholar.sun.ac.za/handle/10019.1/135903 |
| work_keys_str_mv | AT rennerdanielstefan conceptforavirtualassistantwithanllmbasedmultiagentsystemtoprocessuserinputinintralogistics |