Step 1 Odometry update (Prediction step)

Parent: Basic EKF for SLAM

Source: SLAM for Dummies

First step in the three-step EKF

• Update current state using odometry data
• Based on the controls given to the robot
• Calculate estimate of new POSE

Update equation: prediction model  ($x = x + \Delta x \cdot q$)
Or in a simple model, neglect the error term $q$

1. State vector gets updated via the prediction model

2. Jacobian of the prediction model also needs to be updated every iteration (with the controls deltax, …)

3. Process noise Q updated to reflect control terms

4. Calculate new covariance for robot POSE (onlt the top left of the P matrix using the Jacobian of pred. model A and process noise Q

5. New cross correlations in  P matrix (top three rows)

6. Transpose the cross correlations to the leftmost three columns

Due to odometry errors, this estimate is not exact.