Thesis Worker - Finding representative roads of a vehicle's missions

Background

As the Earth becomes warmer, many legislative and executive bodies have set carbon dioxide emission goals that are supposed to reduce and eventually reverse global warming. To measure whether these goals are reached, a method for quantifying CO2 emissions has been developed by the EU Commission called VECTO. It simulates the vehicle motion corresponding to a set of driving cycles (≈speed profiles) that is said to represent driving of most heavy vehicle operations, neglecting the differences in vehicle use of specific customers. An alternative to this has been developed at Chalmers, where the road, weather, mission, and traffic characteristics are described using stochastic processes. These models describe the environment that a customer is operating in and can hence be used for a multitude of studies, including, for instance, quantifying energy consumption and CO2 emissions or setting performance requirements on new vehicles. Arriving at such conclusions has historically been done using repeated simulations (MCM), but a solution was recently found by solving the stochastic differential equation (SDE) that governs the vehicle motion.

Problem motivating the project

A final confirmation of results is done through real-world testing. In the ideal scenario, each vehicle model shall be tested on all roads, weather conditions, traffic densities, and cargoes. This is unfortunately not possible, and the testing is instead conducted over shorter test routes. But how can one be sure that the test route is representative of the vehicle use of a specific customer? And where should one conduct the test drives to comply with specific customers’ vehicle use?
To answer these questions, the students may want to try the following:

• Given a large set of vehicle log files from one customer, derive the stochastic processes, specifically the parameters of these processes. We call this set of parameters: generic stochastic operating cycle (sOCg)
• Assume that the test vehicle starts at point A, and the driver can go left, right, and continue at each intersection. Find the sequence of actions that makes the stochastic parameters of the test route converge the fastest to the sOCg, for instance, using grid search algorithms, maximum likelihood, and methods from stochastic calculus.

Research Question

1. How can road information, such as speed limits, curvature, etc, that are affecting the energy consumption of road vehicles be modeled stochastically, and how can one estimate the parameters of these models from measurement data?
2. Using stochastic models, how can one find a route that shares the same characteristics as the sOCg?
3. How can one estimate the stochastic parameters of a collection of transport operations, i.e, create, motivate, and define the sOCg?

Requirement on student background:

The project is suitable for 2 students with a background in mathematical modeling, stochastic calculus, optimization theory, and data analytics. Examples of Master’s programs at Chalmers are MPSYS,MPCAS and MPENM.

Supervision and examination:

• Carl Emvin, (Academic supervisor), Chalmers University, Phone: +46 31 772 2313
• Fredrik Bruzelius, (Examinator & academic supervisor), Chalmers University, Phone: +46 31 772 3628
• Rickard Andersson

Administrative

Suitable number of students: 1-2
Number of credits: 30 credits
Starting date: 15th January
Stakeholder: Volvo Group
Application to: https://jobs.volvogroup.com/?locale=sv_SE
Physical location: Volvo Group (CampX)

Last application date: 12th of December.