For a century has Internal Combustion Engine Vehicles (ICEVs) been the dominating means of road transportation, but the era of Electric Vehicles (EVs) is finally approaching. As a result, new challenges arise, amongst which a rather problematic challenge is range anxiety. The problem stems from the lack of
charging infrastructure, the low power delivery at the connector, and yet small battery sizes. Until the infrastructure and technology have matured, range anxiety is difficult to avoid for the drivers. However, the fear of misfortune can be reduced by providing the driver with a robust and accurate measure of residual
driving range. That is the problem to be solved by projects such as the U-FEEL project . To predict the residual driving range, several strategies can be deployed including model-based approaches such as determining the average specific energy consumption (SEC) and model-free approaches such as radial bias function neural networks (RBF NN). No matter which strategy is deployed, they all share the need for road data to make a good prediction. Some information is difficult to come over such as vehicle speed, but can be estimated by a driver model that is aware of speed signs, traffic, etc. The information required to estimate the residual range is packed into a Chalmers format called deterministic Operating Cycle (dOC), and it includes; slope, curvature, speed limit, etc., see table 1.
Figure 1: The deterministic Operating Cycle (dOC) format.
Such data are available by 3rd party data companies such as; HERE road maps, Google Maps, Klimator, and more. These companies provide routing services that allow the user to request data from their Application Programming Interface (API). The user typically does a post request to the API containing information about the vehicle together with a GNSS-body. The service then calculates the most probable route and sends back a JSON object containing the road segments that the vehicle has traveled on containing the data requested by the user. To use the data in Complete Vehicle Simulation tools like VEHicle PROPulsion (VehProp), a conversion from JSON to a compatible format such as the dOC format has to be performed. The conversion is not trivial, for example, the weather is a function of time and space, but should be represented by one dimension of space, the distance traveled. A schematic representation of the workflow of a vehicle engineer running simulations is presented in Figure 2.
Figure2: High-level representation of typical workflow of complete vehicle simulation.
Blue arrows are the focus area of your thesis.
In a previous project COVER , dOCs were created using a combination of road data, vehicle log data, and weather data. In the merge of all this data, it becomes crucial to have a good distance estimate since all the information is represented as a function of distance in the dOC format. The vehicle log data
can be represented as a function of the odometer measurement which is rather accurate. The data from APIs (road maps data and weather data) contains coordinates generally in the format WGS84 and can be mapped to a specific cartesian coordinate at the world’s surface with some accuracy. Using haversine
or Vincenty’s formulae, one can estimate the path length, but how accurate is that estimation? how much can the error be reduced if one were to use other methods like NURBS length estimation? Is the error acceptable? etc...
- Fluent in Python.
- Have experience in working with APIs.
- Have a good understanding of Databases and know how they function.
Description of thesis work
The core of this study is to investigate and quantify the accuracy of distance calculations using different Coordinate systems and estimation methods. To do so, one should first create the back-end as depicted in Figure 3 and after that sort and process the data into the simulation-friendly format dOC. The student
will first have to study the different APIs available and choose the most suitable API concerning the application. The student should then deploy a number of path length estimation methods and compare their accuracy. The student is recommended to quantify the accuracy using standard measures but also to showcase the performance of each method in a case study.
Figure3: Schematic representation dOC creation using 3rd party data.
The research question is the following:
How is the path length of a road accurately estimated? How do different methods for path length estimation perform for short and long paths? What triggers a large error in the different estimation methods?
- A backend that collects road data.
- A implementation of path length estimation using Haversine, Vincentys or another method (maybe NURBS length estimation).
- An accuracy comparison of the different methods used.
- A study in when the different methods are error-prone.
- Literature study
- Code the backend, requesting data from the API of your choice.
- Calculate path length.
- Merge the responses into dOC files.
- Compare and evaluate the different path length estimation methods used.
 Chalmers University of Technology, Nytta och förtroende för elektriska fordon U-FEEL. [Online]. Available: https://research.chalmers.se/en/ project/?id=11005.
 Chalmers University of Technology, COVER – Real world CO2 assessment and Vehicle enERgy efficiency. [Online]. Available: https://research. chalmers.se/en/project/8239.Thesis Level:
15th of January 2024Number of students:
1 - 2 studentsTutor:
Carl Emvin (Chalmers) - email@example.com
Fredrik Bruzelius (Chalmers) - firstname.lastname@example.orgResearch group:
Vehicle Engineering and Autonomous SystemsExaminer:
Ola Benderius (Chalmers) - email@example.comKindly note that due to GDPR, we will not accept applications via mail. Please use our career site. Physical location:
Chalmers M2 and possibly also at Volvo GTT.IMPORTANT NOTE: The Master Thesis is completely handled by Chalmers & Volvo will not cover any remuneration or payment connected to the Master Thesis. Volvo is a partner in the externally funded project U-FEEL.