Classifies longitudinal behavior of vehicles based on the probabilistic output from the constrained Kalman filter.
The algorithm works as following:
- vehicle's longitudinal and lateral position is fed to CACV (Constant Acceleration Constant Velocity) Kalman filter
- resulting estimated states are smoothed with RTS smoother
- new estimations and covariance matrices are used in probability truncation
For usage please download NGSIM dataset (csv measurements on US101) and place it in data folder as measurements.csv
Based on the MATLAB algorithm written by Dan Simon