On April 29th, 2021, Niklas Petersen successfully defended his Ph.D. thesis entitled “Multi-model bus arrival prediction with intelligent handling of uncertainties.“

Congratulations, Niklas!

He was supervised by Professors Francisco C. Pereira and Filipe Rodrigues.

Examiners were Prof. Thomas Kjær Rasmussen (DTU Management), Prof. Oded Cats (TU Delft), and Prof. Bilal Farooq (Ryerson University).

Chairperson at defense was Prof. Anders Fjenbo (DTU Management).

Find below the abstract of the thesis! For more information, feel free to contact Niklas at iklch@dtu.dk.

Waiting for the bus will presumably never be exciting. However, digital departure screens and smartphone apps (e.g. Rejseplanen) now allow the traveler to keep track and know when their bus is arriving. At least that is how it is supposed to work, but in reality, many have experienced a digital departure screen reaching “zero” and yet no bus was there. Behind the scenes, each bus is constantly sending its current location based on GPS to a central prediction system, that tries to guess all the arrivals of each and every bus. But it is a difficult task with varying congestion in the cities, unpredictable traffic signals, road works, and incidents. This Ph.D. firstly explores how machines can model and learn better to guess more accurately using the advancements in computational power and data available. But there is a catch: the more complex machine learning techniques do not always make sense, and they are prone to fail when their underlining assumptions are not met. Therefore, secondly, a multi-model approach is presented where multiple, both simple and complex, machine learning techniques can compete and support each other for enhancing robustness, flexibility, and accuracy in a real-world prediction system. The presented multi-model approach is shown to outperform single-model approaches. The thesis also provides two studies on how the combination of location data from GPS and machine learning technology can be used together to extract valuable information about travel behavior, including an example of measuring the walking time needed when travelers are transferring from a bus to a train in the public transport system.