Based on the multiObjectTracker class definition and its functionality, it is designed to effectively handle asynchronous sensor data, maintaining accurate tracks even when sensors update at different rates. To explore the class definition, selecting the multiObjectTracker class name in any MATLAB code and using Right Click > Open Selection or Ctrl+D provides direct access.
The multiObjectTracker class inherits from matlabshared.tracking.internal.fusion.GNNTracker, which uses the Global Nearest Neighbor approach for assigning detections to tracks. This method remains robust even when update rates vary, as it relies on a cost matrix to associate measurements with tracks.
Track states are predicted to the time of each new measurement using the predictTracksToTime method, ensuring that updates remain consistent with the latest sensor data. The updateTracks method manages the process by predicting track states to the measurement time and correcting them if detections are assigned. Tracks that are not assigned detections are coasted, allowing them to persist without immediate updates.
The ConfirmationThreshold and DeletionThreshold properties help manage track confirmation and deletion, maintaining or removing tracks based on their update history.
In summary, the multiObjectTracker object predicts detection measurements and corrects them with actual detection data as it becomes available. Hence, it can handle sensor data arriving at different rates. Additional details can be found in the documentation by using the following command:
web(fullfile(docroot, 'driving/ref/multiobjecttracker-system-object.html'))
