Cimrman Reconstruction Module

"Algorithm so intuitive, you would think Jára Cimrman designed it."

Cimrman is a track reconstruction module for SuperNEMO. It reconstructs charged particle trajectories using a combination of Legendre transform and maximum likelihood method. Named in tribute to the great Czech thinker Jára Cimrman. The module takes CD bank as an input, reconstructs the trajectories within its own internal structure and outputs both TCD and TTD data banks, meaning it is responsible for clustering of tracker hits and for trajectory reconstruction at the same time. Currently it is able to produce line and polyline trajectories. It also detects ambiguous parts of tracker data and provides all alternative possible solutions (tracker_trajectory_solution and tracker_clustering_solution) in such cases.

To obtain the final PTD bank you then need to apply Charged Particle Tracking module that extrapolates the verteces and creates the PTD bank.

Module installation

In the future the module will be integrated in Falaise by default, but at the moment the module has to be installed independantly. For this reason, I temporarily provided the installation script install.sh

chmod 755 install.sh
./install.sh

Module example usage

cd testing/test2/
flreconstruct -i test_SD.brio -p reco.conf -o test_TTD.brio

Module configuration

See the provided example configuration file testing/test2/pipeline.conf, which contains the default setting.

1. General configuration section:

   • visualization_2D: Creating and saving a png image for each solution of each event.

   • visualization_3D: Creating and saving a 3D object root file for each solution of each event.

     Warning: intended for debugging purposes only, do not use on large datasets unless you want to generate enormous amount of files.

   • electron_mode: "polyline" mode enables polyline trajectory, "line" disables polylines

   • reconstruct_alphas: turns on alpha particle tracking (or electrons without a reference OM hit)

   • use_provided_preclustering: use (pre)clustering from existing TCD bank if available (for example pcd2cd output)

   • force_default_sigma_r: not using the drift radii uncertainties provided by Falaise and using default value instead

   • default_sigma_r: universal default uncertainty for drift radii

2. Electron clustering section:

   • save_sinograms: saving the images of sinograms produced during Legendre transform based clustering.

     Warning: intended for debugging purposes only, do not use on large datasets unless you want to generate enormous amount of files.

   • max_distance: the maximum distance between two groups of tracker hits to be still considered as a part of one cluster distance between tracker hits i and j = distance of their anodes: D(i,j) := sqrt( (x_j - x_i)² + (y_j - y_i)² ) distance between two groups of tracker hits G₁ and G₂ = minimum distance between among all possible pairs of tracker cells from the two groups) D(G₁, G₂) := min{ D(i,j) | i from G₁, j from G₂ }

   • hit_association_distance: clustering first finds a rough estimate for a line fit. Then associates tracker hits to it based on this distance threshold

   • no_iterations: number of iterations of the itterative zooming grid search algo used to find the maximum of the sinograms

   • resolution_phi: number of bins used for the grid search algo. starting range for phi is (0, phi)

   • resolution_r: number of bins used for the grid search algo. starting range for r is optimized to be minimal for each event.

   • max_initial_precision_r: If the initial range for r is low enough, number of used bins in r is reduced. In that case one bin = clustering_max_initial_precision_r

   • zoom_factor: each iteration of the search algo zooms in with this factor to achieve better precision

   • uncertainty: defines the blurring of the sinograms to compensate for errors and uncertainty caused by scattering

3. Alpha clustering section:

   • clustering_resolution_phi: number of bins used for the alpha cluster estimation - provides angular bounds for the alpha track
   • save_sinograms: saving the images of sinograms produced during Legendre transform based clustering.
   • min_possible_drift_time: minimum allowed value of drift time
   • max_possible_drift_time: maximum allowed value of drift time
   • time_step: size of the time step - drift radii are recalculated and the Legendre transform based track estimation is done for each time value
   • phi_step: size of the angular step used for the Legendre transform based track estimation
   • max_r: the maximum considered distance between the track and the center of the cluster
   • resolution_r: number of bins used to investigate the [-max_r, max_r] range
   • uncertainty: defines the blurring of the sinograms to compensate for errors and uncertainty caused by scattering

4. Polyline reconstruction section:

   • max_extention_distance: trajectory can be extended by associating additional tracker hit that satisfy the "clustering_hit_association_distance". This parameter defines the maximum possible elongation of the trajectory

   a) Finding sharp kinks - only vertical part of segments is changed

   • max_vertical_distance: maximum possible vertical distance of the two segments to be connected (computed in the point of their intersection in the horizontal projection)
   • min_tracker_hits_distance: tracker hits on both segments must be at least this close to the candidate kink to prevent the creation of distant "stranded" fake kinks
   • max_kink_angle: limits the 3D kink angle for the kink finder algo
   • min_distance_from_foil: prevents the kink to be created close to the foil

   b) Finding tiny kinks - all parameters of segments can be slightly changed

   • max_trajectories_middlepoint_distance: maximum allowed horizontal shift of the segments to be connected into a polyline
   • max_trajectory_endpoints_distance: maximum allowed 3D distance for two endpoints of two trajectories to be connected into polyline
   • max_trajectory_connection_angle: maximum alowed 3D angle for two trajectories to be connected into one polyline trajectory

Software design

The data structure is build in several steps of the reconstruction as suggested in the scheme bellow. The colorcoding indicate the steps responsible for the creation of the obejcts.

You can find detailed description of the algorithm and the derivation of all methods in my Master thesis. Specifically, Chapter 5 describes the software structure of the algorithm in more detail.