Tasks

[maxschnettler]

• internal status variables + hyperparameter create list of all variables and their specifications numpy-arrays vs. class objects? --> research about computation efficiency

• integration of configuration files research about config files and their integration in Python

• pointcloud example: create / integrate small point cloud example as functional test for further development

• implementation of filter logic spawn new targets prediction correction

[maxschnettler]

Required variables

Hyperparameter

• max. number of tracks
• max. number of mixture compontents per track
• survival probability (const. at first) --> death probability = 1 - p_s
• birth probability
• detection probability
• clutter density kappa
• pruning threshold for existence probability
• process noise covariance Q
• observation noise covariance R
• pruning threshold for association hypothesis weights
• mixture components merge threshold

Problem config parameter

• problem dimensions / number of states
• motion model (e.g. Constant Velocity Model) --> F and H

Internal state variables

• target class (existence probability, label, phd/pdf)

[maxschnettler]

Class structure:

target class contains existence probability, label, (motion model), and pdf representation/object

lmb logic prediction and correction step of lmb, including association and sampling

gaussian mixture pdf

[phquentin]

We may take Murtys Algorithm from here https://github.com/abewley/sort/blob/master/sort.py

[phquentin]

Hungarian Algorithm: https://github.com/xinshuoweng/AB3DMOT/blob/master/evaluation/munkres.py