Covariance matrix P
Source: SLAM for Dummies
s. also EKF matrices
Covariance matrix P
- Covariance: measure of correlation of two variables
- Correlation: measure of degree of linear dependence
| A | covariance of the robote POSEupdated in Step 1: Odometry update | 3x3 |
| B .. C | covariance on the first .. nth landmark Step 3: New landmarks | 2x2 |
| D | covariance between POSE and first LMupdated in Step 1: Odometry update | 2x3 |
| E, etc | E = D^T, etcupdated in Step 1: Odometry update | 3x2 |
| F=G^T | Step 3: New landmarks |
Initially $P = A$ (robot has not seen any LMs)
Include initial uncertainty
- There will often be a singular error if the initial uncertainty is not included
- good idea to include some initial error even though there is reason to believe that the initial robot position is exact
- initialise default values for the diagonal
Top three rows: cross correlation between POSE and landmarks updated in step 1 (odometry update)