Extended Kalman Filter Project

The goal of this project is to implement an Extended Kalaman Filter using Radar and Lidar measurements.

Project Structure

The crux of the code is implemented in 3 files:

• FusionEKF.cpp
• kalman_filter.cpp
• Tools.cpp

FusionEKF: This is the module that is responsible for initializing and calling the kalaman filter prediction and update functions based on the type of the input measurement. The inpupts can be either from a Radar or from a Lidar

kalaman_filter.cpp: This is the module that implements the kalman filtering prediction and update equations - both the linear case(Lidar) and the non linear case (Radar)

Tools.cpp: This module implements the RMS calculation as well as the Jacobian matrix calculation that is required for the non-linear update equations

Results

For the dataset under Data folder, the RMSE values when using both Radar and Lidar inputs are given by:

Variable	RMSE
px	0.0973178
py	0.0854597
vx	0.451267
vy	0.439935

Experiment with using only one of the sensors

In this experiment, only one of the sensor measurements was used as input to the Kalman Filter. The RMSE results are given below.

Usng only Lidar measurents:

Variable	RMSE
px	0.183795
py	0.154202
vx	0.605092
vy	0.485836

Usng only Radar measurents:

Variable	RMSE
px	0.233172
py	0.335998
vx	0.617771
vy	0.678604

We can see that the Radar by itself gives the least accurate estimates. Using just Lidar gives better accuracy in the final measurements. However, fusing information from both the sensors gives the best results.