Track finding rework

[paulgessinger]

Problem

The current CKF is unable to find inwards from the seed location. This can be badly fixed by propagating to the IP and then starting the search but we need to work out a comprehensive solution.

Part 1: Improve current CKF

• Run inwards filtering from seed with restricted branching
• smooth forward to seed location
• run forward search with branching
• combine track states to complete tracks (when we copy tracks after selection)

Action items

• #2317
• #2318
• #2319
• [ ] How do we terminate the backward search? Currently uses the IP as a target surface during filtering

Part 2: Implement alternative track finder based on ATLAS strategy

[github-actions[bot]]

This issue/PR has been automatically marked as stale because it has not had recent activity. The stale label will be removed if any interaction occurs.

[paulgessinger]

@XiaocongAi suggests we could implement early rejection of measurements by only considering measurements inside a certain cartesian distance from the predicted parameters. This could save us a few $\chi^2$ calculations, especially on low-radius layers.

During the discussion, we were wondering if it might be possible to run branching track finding inwards and outwards, then take the cross-product between all endpoints, and just look at some compatibility metric between smoothed parameters on the same reference surface. Not sure if this is good enough.

[XiaocongAi]

@XiaocongAi suggests we could implement early rejection of measurements by only considering measurements inside a certain cartesian distance from the predicted parameters. This could save us a few χ2 calculations, especially on low-radius layers.

During the discussion, we were wondering if it might be possible to run branching track finding inwards and outwards, then take the cross-product between all endpoints, and just look at some compatibility metric between smoothed parameters on the same reference surface. Not sure if this is good enough.

To avoid additional inwards and outwards compatibility check, we can first propagate outwards, and then propagate inwards using the smoothed track parameters at the first measurement with the outward propagation.

[XiaocongAi]

Regarding the 'How do we terminate the backward search? Currently uses the IP as a target surface during filtering', can't we terminate when we reach the boundary of volume for the beampipe?

[andiwand]

Regarding the 'How do we terminate the backward search? Currently uses the IP as a target surface during filtering', can't we terminate when we reach the boundary of volume for the beampipe?

for this purpose I implemented a filtering target surface which could be the perigee for examples. thinking of cosmics or particles that appear inside the detector I think terminating on the beampipe volume might be insufficient

https://github.com/acts-project/acts/pull/2319

[github-actions[bot]]

This issue/PR has been automatically marked as stale because it has not had recent activity. The stale label will be removed if any interaction occurs.