Master Thesis: Learning time-series embeddings for auxiliary energy consumption of HD BEVs

Job description

Transport is at the core of modern society. Imagine using your expertise to shape sustainable transport and infrastructure solutions for the future? If you seek to make a difference on a global scale, working with next-gen technologies and the sharpest collaborative teams, then we could be a perfect match. 

 

Background:
The Virtual Product Development and Digital Services (VPD & DS) team within Volvo GTT is focused on delivering quality products based on fact-based decisions and a data-driven approach. The team aims to improve the quality of the products, reduce development costs, and enhance customer experience through its connected services.

A key feature for both sales and customers is the ability to predict energy consumption for our heavy-duty vehicles. Time-series sensor data is part of the logged data in vehicles. Useful embeddings can be extracted from multi-variate time-series data and then be utilized for downstream tasks, for example, data-driven model for predicting auxiliary energy consumption. As a result, this would enhance the accuracy of total energy consumption predictions for heavy-duty vehicles.

We believe you:
•    Are a student studying an MSc program in computer science, machine learning or equivalent.
•    Feel confident in programming in Python and/or PySpark, preferably with some experience in relevant frameworks such as PyTorch, Keras, or Tensorflow.
•    Have strong knowledge in deep learning methods, e.g., transformer models; practical experiences in applying machine learning techniques on real-world problems is a merit.
•    Having worked with streaming data, multi-variate time-series data, understanding vehicle usage and energy consumption is a merit.
•    Having worked with time-series foundation models and curriculum learning is a merit.

Thesis Description:
As a master thesis student, you will be working with test or customer data of Volvo trucks to study the embeddings for time-series data. Several embedding models/methods will be explored, evaluated, and compared. You will work closely with the Advanced Analytics Team, and have the opportunity to collaborate with other domain experts as well as various stakeholders in different tech streams. In addition, you will be part of a highly motivated team working with data-driven methods for predictive maintenance, increasing the uptime of heavy-duty vehicles.
In this thesis titled, “Learning time-series embedding for analyzing auxiliary energy consumption of HD BEVs”, you will be working with multi-variate time-series data, collected with on-board sensors from heavy-duty vehicles. One important aspect is to learn useful embeddings for time-series analysis. In addition, you will adapt promising methods using self-supervised learning for time-series embedding. You will validate and evaluate the proposed approach for relevant downstream tasks, such as forecasting energy consumption, clustering transportation types, analyzing usage profiles, and activity recognition. 

Furthermore, machine learning models will be built for forecasting auxiliary energy consumption based on the learned embeddings, grouped by identified clusters. The performance and modularity of the code should also be a major focus of this thesis. As a stretch scope, you will explore the time-series foundation models with the embeddings, e.g. auto-former, informer, PatchTST, etc. Generation of synthetic data is an extended task as well.

Objectives and learning outcome:
•    Literature survey including review of related machine learning methods, in particular for applications addressing time-series data.
•    Develop and implement the ML models for time-series embedding
•    Compare and evaluate ML models, for downstream tasks.
•    Write a Master Thesis report and present the results at the company.
•    Upon successful completion of this work, a feasibility study for further research on this domain is desirable.

Duration:
•    The duration of the Thesis is 20 weeks (Approx.).
•    The thesis starts in January 2025 or earlier if possible.
•    30 ECTS (academic credits) if in agreement with your Thesis Advisor in the University.
•    This thesis is suitable for 1-2 students; 
•    On-site presence is preferred and appreciated, as it is easier to collaborate with our teams, as well as getting access to various resources. 

 

Last application date is 26th of November. If you have questions please contact  zhenkan.wang @volvo.com

We value your data privacy and therefore do not accept applications via mail. 

 

Who we are and what we believe in 
Our focus on Inclusion, Diversity, and Equity allows each of us the opportunity to bring our full authentic self to work and thrive by providing a safe and supportive environment, free of harassment and discrimination.

 

Group Trucks Technology are seeking talents to help design sustainable transportation solutions for the future. As part of our team, you’ll help us by engineering exciting next-gen technologies and contribute to projects that determine new, sustainable solutions. Bring your love of developing systems, working collaboratively, and your advanced skills to a place where you can make an impact. Join our design shift that leaves society in good shape for the next generation.

Job Category:  Advanced Analytics

Organization:  Group Trucks Technology

Travel Required:  No Travel Required

Requisition ID:  15635