Master Thesis Student

MASTER THESIS PROPOSAL | Traton

How Much Data Is Enough?

Optimal data sources and sampling strategies for energy consumption estimation of connected electric vehicles

Background

Estimating how much energy an electric vehicle consumes and how its state of charge (SoC) evolves depends on the availability and quality of vehicle data, as well as the frequency at which it is collected. At lower sampling frequencies, fewer observations are available to capture changes in vehicle state, which can reduce estimation accuracy. Selecting suitable data sources and sampling frequencies is therefore central to developing reliable predictions. Even where European regulations or standards apply, access to relevant data may remain limited. Many OEMs share only a small portion of their proprietary telematics data, often at very low and irregular frequencies. External devices such as OBD dongles can provide additional or more frequent measurements, but they add hardware, installation, and maintenance costs. Collecting data more frequently can also increase the costs of transmitting, storing, and processing signals across a fleet.

Problem statement

The objective of this work is to evaluate how accurately vehicle energy consumption can be estimated from possibly sparse available data at different sampling frequencies. The study will quantify how estimation accuracy changes according to varying sampling frequency and availability of data. Resulting estimates will be compared against the vehicle’s measured energy consumption, which uses higher-frequency measured data. As a secondary consideration, the accuracy gains obtained by using more frequent data collection will be assessed to understand if the use of external measurement devices is justified. The expected outcome is an evidence-based analysis on suitable data sources and sampling frequencies for scalable connected-vehicle applications as well as a cost analysis of using external devices for increased accuracy.

Research questions

• How accurately can energy consumption and SoC be estimated using only OEM data, without external devices?
• When is an external device justified? When does the gain in accuracy outweigh its cost? What are the required specifications that an external device must meet to allow effective data collection and sampling?
• How much does estimation performance improves with an external, tunable device compared to using OEMs’ existing solutions? Quantitively study the benefits of the different options.
• What is the best tradeoff between sampling frequency and estimation performance? What is the best strategy to support an economy of scale?
• How can an adaptive strategy be implemented? How well does it perform?

Possible approaches

The topic can be approached in a more theoretical or a more practical way, depending on the student’s interests:

• Signal processing: spectral content of driving and battery signals, Nyquist and aliasing, down sampling, reconstruction and irregular sampling, information-theoretic measures.
• Modelling: physics-based longitudinal vehicle and battery models compared with data-driven models; how well model parameters can be identified at different sampling rates.
• Estimation: Kalman filters and state observers working with sparse, delayed or asynchronous measurements.
• Cost–benefit analysis: accuracy compared with storage, bandwidth, and computational costs, presented as a Pareto front of possible data strategies.
• Validation: high-frequency reference data that is artificially down sampled, compared with real low-frequency OEM data.

Expected outcomes

• An analysis showing how estimation error varies with data source and sampling frequency, supported by graphs and curves.
• Evidence-based recommendations on when an external device is warranted and which sampling rates to use, including an evaluation of adaptive (high-then-low-frequency) strategies. Where useful, results may also be presented as a Pareto front to support decision-making.

Your profile

• MSc student in Electrical/Mechanical/Automotive Engineering, Applied Mathematics, Data Science or similar.
• Programming experience in Python and/or MATLAB.
• Interest in signal processing, estimation and control, or vehicle dynamics.

Contact persons and supervisors

Boretti Tea tea.boretti@scania.com

Application:

Your application must include a CV, personal letter and transcript of grades

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.