Tracking particles at the High Luminosity LHC accelerator will be a major challenge. In short, for each collision of two bunches of protons, approximately 100.000 3D points have to be associated into 10.000 trajectories, approximate arc of helices. The Tracking Machine Learning challenge (TrackML) took place on Kaggle from May to August 2018 and on Codalab from October to March 2019. A wealth of track pattern algorithms have been exposed : some pure combinatorial using trajectory following, some assisted with Machine Learning, some with unsupervised clustering. Given that this was a competition, the ultimate algorithm is most likely a combination of good ideas (some being very original) used by different competitors. Acts is the C++ tracking tool suite which has been used to simulate the TrackML dataset. This versatile framework also allows reconstruction algorithms to be run. The goal of this project is to examine a number of TrackML submitted algorithms, to port the most promising ones (two or three) to Acts and to build an optimal one using ideas exposed in the two phases of the challenge.
A suite of efficient tracking reconstruction algorithms implemented in Acts framework
Experienced with C++ and python