Kalman Review MKII¶

• 10:00 – 12:00 Norton Room
• 13:00 - 17:00 R76 G06

See also #1251 and previous meeting

Overview¶

Extrapolate/Propagate ==

• UpdatePropagator - sets up transfer function with sites information
• CalculatePredictedState - extrapolate state vector
• CalculatePredictedCovariance - extrapolate covariance matrix
• SubtractEloss - take off energy
• addMCS - add noise

Modified class structure

• Kalman Seed
• Kalman TrackFit
• Kalman Propagator
• Kalman Filter

General discussion¶

ISSUE: KalmanSite should be taken out of Track
Solved

ISSUE: MiceModules implementation is rotated through 180 degrees

• ACTION: Adam to fix MiceModules implementation
• ACTION: Adam to fix PR and MC, MC Digitisation issues arising
• ACTION: Ed to fix Cluster finding SP and Kalman

KalmantrackFit¶

ISSUE: Particle mass constants should be in Utils/Constants.hh (new class)
Solved. But not a new class, just constants stored in a namespace.

ISSUE: KalmanSeed ComputeInitialStateVector should use natural G4 units
Solved

ISSUE: Filter should take dimension from the seed not hard coded
Solved

ISSUE: ComputeChi2 hard coded n_parameters
Solved

ISSUE: initialise propagator, filter means delete filter, propagator
Solved

ISSUE: Extrapolate - still needs to throw an exception if int i out of range
Now accessing vector elements with std::vector::at instead of [] operator - does the required bound checking.

Kalman Propagator¶

ISSUE: KalmanSite should be clearer named KalmanState
Solved

ISSUE: _F_transposed is not member data, should be F_transposed
Solved

ISSUE: _mass should be taken from Utils/Constants.hh
Solved

ISSUE: beta2 should be beta*beta etc
Solved

ISSUE: 28.816 should be a constant
Solved

ISSUE: muon_mass should not be 105.7
Solved

ISSUE: 13.6 should be at least a constant; MCS formula should be a member function
Solved

ISSUE: UpdateBackTransportationMatrix Remove temp
Solved

ISSUE: remove compiler warnings
Solved

ISSUE: CalculatePredictedState multiplication is a bit weird
Solved

KalmanFilter¶

ISSUE: could store sqrt(12)
Rejected

NOTE: your code can't handle misalignments
Solved by recovering some code removed after the first code review.

Data structure¶

ISSUE: Make sure you do deep copies
Solved

ISSUE: Comment up data members
Solved

Tests¶

ISSUE: Tests - recommend for each public function write a test

ISSUE: Thesis chapter should be merged into the latex documentation when its ready with a few modifications to reference against e.g. class structures and call diagrams

Notable issues from the previous meeting:¶

ISSUE: No scheme for chi2 when we have multiple adjacent channel hits - we took note but decided this was an action for tracker group (not necessarily Santos)

ISSUE: Need to remove Monte Carlo data from the tracker recon data structure - predicated on #1197 which was moved to Taylor

ISSUE: Track object should point back to the clusters that were use to construct it, using a TReference. We noted that the rest of the tracker data structure also needs to use TReferences for cross links. Ian Taylor to advise.

ISSUE: Noted that _mass assumption of muon mass may bias the pid measurement, i.e. may expect better fits for muon assumption even when we have pions because the seed contains muon assumption. Issue to be studied by global group when they have pid algorithms.