Simulate a V2I environment by generating noisy measurements from 3 separate radar towers that will return the distance of the vehicle from each radar tower at each time step. These noisy measurements will then be propagated through a KF to estimate the current state of the vehicle in the 2D environment and then compare the results to the ground truth. I will assume no knowledge of the vehicle motion (random walk motion model) and start with an initial guess for the state.
- Open Matlab and load ground truth tragectory 'transformations>07.mat' whcih was taken from KITTI data set and parsed to generate transformations as well as ground truth data for map sequence 07 (http://www.cvlibs.net/datasets/kitti/eval_odometry.php)
- Ensure "generate_measure.m" and "Ali_V2I_run.m" are in the same path
- Run "Ali_V2I_run.m"
The 2D ground truth trajectory of the 07 KITTI data sequence, as well as the ๐ฅ,๐ฆ coordinates for the 3 radar towers are displayed below.
Ground Truth Trajectory & Radar Tower Positions
Kalman Filter Approximation va Ground Truth
Deviation from ground truth for x, and y coordinates
Visually the Kalman Filter approximation mirrors the trajectory reasonably well with some noisy state approximations that result in oscillations away from the ground truth. The deviation of the state estimate vs the ground truth for the x coordinate as well as the y coordinate. It is apparent that the deviation error remains within the 3-sigma contours evaluated.