Master Thesis

Using Large Language Models (LLM) for Transport Mission planning optimization of BEV fleet with trailer units


Background of thesis project


In an era where sustainable transportation solutions are paramount, the electrification of commercial vehicles has gained considerable attention. This master's thesis delves into the realm of Battery Electric Vehicle (BEV) trucks and their trailers, focusing on optimizing routing, charging, loading, and unloading processes. In stark contrast to classical optimization methods, which often struggle with the intricacies of BEV logistics, this research leverages the power of Large Language Models (LLM) like ChatGPT. By harnessing LLM's adaptability and advanced problem-solving capabilities, this study aims to revolutionize BEV truck operations, ushering in a new era of efficiency and sustainability in the logistics industry.


The article [1], from Google Deep mind, describes a novel approach called Optimization by PROmpting (OPRO) that utilizes large language models (LLMs) as optimizers for solving various problems. OPRO works by describing the optimization task in natural language and having the LLM generate new solutions based on previously generated ones. The study demonstrates the effectiveness of OPRO on tasks like linear regression, traveling salesman problems, and prompt optimization. The results show that OPRO can outperform human-designed prompts by significant margins on different tasks. This work highlights the potential of using LLMs for optimization tasks and their superior performance compared to traditional methods.
Suitable background
Good knowledge of Control theory and optimization
Good mathematical skills and interest in Artificial general intelligence (AGI)

Description of thesis work
The purpose of this master thesis contains three parts:
Setup optimization of classical optimization problems such as salesman problem using OPRO approach with LLM models that are free to use for research and commercial use, such as LLAMA2 from meta [2].
Automate prompting to find solutions and automatically checks if the solution is feasible, i.e. fulfilling optimization constraints and how optimal are the solutions found.
Setup the more complex problem of transport mission planning and execution by minimizing total cost of operation for truck BEV fleet with closed system of trailers.

The thesis work will include control theory and optimization. The work will be carried out at Volvo Group Trucks Technology. The thesis is recommended for one or two students per team with good mathematical skills and interest in Artificial general intelligence (AGI). AGI is the representation of generalized human cognitive abilities in software so that, faced with an unfamiliar task, the AGI system could find a solution.


References:
[1]. Chengrun Y., Xuezhi, W., Yifeng, L., Hanxiao, L., Quoc V.L., Denny Z., and Xinyun C., “LARGE LANGUAGE MODELS AS OPTIMIZERS”, ArXiv, September 2023, 2309.03409.pdf (arxiv.org)
[2]. Llama2, Large Language Model, Meta, source: https://ai.meta.com/llama/, (September 2023)


If you find this proposal interesting send your application with CV and grades to:
dhasarathy.parthasarathy@volvo.com


Thesis Level: Master

Language: English


Starting date: Jan 2024.


Number of students: 1 or 2 students per team


Tutor
Contact persons:
Dhasarathy Parthasarathy – Volvo GTT
tel: +46 31 323 5311
mail: dhasarathy.parthasarathy@volvo.com


Jonas Hellgren – Volvo VAS
tel: +46 73 9024761
Mail: jonas.hellgren@volvo.com


Leo Laine – Volvo GTT
tel: 031-322 3805
mail: leo.laine@volvo.com