Master Thesis - Deep Learning for AGV-system design!
Kollmorgen Automation AB / Datajobb / Mölndal
Observera att sista ansökningsdag har passerat.
Visa alla datajobb i Mölndal,
Göteborg,
Partille,
Härryda,
Kungsbacka eller i
hela Sverige Visa alla jobb hos Kollmorgen Automation AB i Mölndal,
Göteborg eller i
hela Sverige About the project
An Automated Guided Vehicle (AGV) system is a fleet of mobile robots that automatically transport goods in a network of fixed virtual roads, designed according to the specification of the warehouse. This network must have the capacity to handle the intended traffic load without traffic jams, which is verified by running simulations. However, such simulations are time consuming, and since the design process is an iterative one, faster feedback directly translates to decreased cost. In this master thesis project, you will research how Machine Learning can be used to speed up the design process by predicting the consequences of different design choices.
Goal of the thesis
You will be given a labelled dataset with many variants of the same intersection, and your job is to develop and train a machine learning model that can predict how well each variant performs. For each variant, a simulation has been run which measures the throughput, i.e. the rate at which goods can be delivered, and this is what you will try to predict.
In the second phase of the project, you will generate new datasets for other kind of intersections. Here, the research question is whether the model can generalize to scenarios that were not in the training data.
By exploring this field of research, you will lay the foundation of new technology that will enable us to reach our goal to install more AGV systems faster
Method
The task is a form of supervised learning, and you will identify and evaluate a range of different models. Relevant metrics such as precision and recall should be calculated for each model, and you shall also explore how the choice of hyperparameters affect convergence, accuracy, and robustness. Roughly, the work will be carried out as follows
• Literature review to identify relevant models. One type of model we want you to study in particular is Convolutional Graph Networks
• Implement and train relevant models on existing data set using python libraries such as numpy, pandas, matplotlib and pytorch.
• Compare the results from the different models and give recommendation on model selection
• Generate new dataset. You will be responsible to select suitable scenarios in dialogue with domain experts at Kollmorgen, and you will use various existing python scripts to generate synthetic data, including labeling. In other words, you will not have to annotate any data manually in this project.
• Repeat training and analysis on new dataset, including testing how well the models generalize to unseen scenarios
About you
This project will require a solid background in mathematics, statistics and programming, and we are therefore primarily looking for master students from the engineering programs, such as Computer Science, Applied Physics, Applied Mathematics and similar.
However, if you are interested in Data Science and Artificial Intelligence, know python (or similar) and are not afraid of eigenvalues, we think you can take on this project if you have the right mindset.
Number of students
2
Thesis Level
Master
Language
Thesis is to be written in English.
Starting date
January 2023
Location
Kollmorgen Automation Office, Mölndal, Sweden
Other information
We will provide computers and you will get great support from your supervisor and other colleagues!
All students are compensated after the thesis is submitted and the final work meets the quality standard that is generally accepted by the university and presented at Kollmorgen.
Send in your application by submitting a CV and/or information about your studies and a short sentence about why you find this project interesting. If you get selected, the next step will be to have discussions with us to finalize the idea. Selection takes place on an ongoing basis, welcome with your application!
Some of the good things that we offer
Innovation Days every twelve weeks - a 24-hour event for the whole company to dig deeper, explore new areas and solve problems!
Gym at the office filled with machines for strength and fitness that is always open and free to use when you need a break in your thesis work
A bicycle storage and free car parking outside our office
Be a part of our ordinary events such as monthly breakfast meetings, Day of Caring (every year we take one day to clean the west coast beaches together) etc
Good fair trade coffee (or tea) with fresh milk (every student who has been up late to study for an exam know how important coffee is)
Fresh fruit if you get hungry
Contact information
If you have any questions about the thesis, you are welcome to contact Rasmus Åkerlund at
rasmus.akerlund@kollmorgen.com.
Check out more open positions (
https://career-agv.kollmorgen.com/jobs), follow us on Facebook (
https://www.facebook.com/kollmorgen.ndc), get to know us better through our Candidate Blog (
https://kollmorgenagv.teamtailor.com/blog), and see videos about our products at our Youtube channel! (
https://www.youtube.com/user/ndcsolutions)Varaktighet, arbetstid
Heltid No Employment but Thesis Project
Publiceringsdatum2022-10-28ErsättningLön enligt överenskommelse
Så ansöker duSista dag att ansöka är 2022-11-30
Klicka på denna länk för att göra din ansökanFöretagKollmorgen Automation AB
Omfattning Detta är ett heltidsjobb.
Arbetsgivare Kollmorgen Automation AB (org.nr 556114-2778),
http://kollmorgen.com/agv Kontakt Malin Pajnert
malin.pajnert@kollmorgen.com 0732 57 12 00 Jobbnummer 7119558
Observera att sista ansökningsdag har passerat.