When testing an error state extended Kalman filter for navigation using IMU and DVL, there are several sanity checks that can be used to ensure that the filter is working correctly. Some of these checks include:
Consistency checks: The filter should be tested for consistency by comparing the estimated state and covariance values with the measurement data. If the estimated state values are not consistent with the measurements, it may indicate a problem with the filter.
Convergence tests: The filter should be tested to ensure that it is converging to a stable solution. This can be done by examining the evolution of the filter's state and covariance over time. If the filter is not converging, it may need to be adjusted or tuned.
Residual analysis: The filter's residual values (i.e., the difference between the measured and predicted values) should be analyzed to ensure that they are normally distributed and have a mean of zero. If the residuals are not normally distributed or have a non-zero mean, it may indicate a problem with the filter.
Outlier detection: The filter should be tested to ensure that it can handle outliers and noisy data. This can be done by intentionally introducing outliers or noisy data into the measurement data and observing the filter's response.
Real-world testing: The filter should be tested in a real-world environment to ensure that it is working correctly under real-world conditions. This can involve testing the filter in a variety of different scenarios and environments, such as in different weather conditions, with different types of obstacles, and so on.
These are just a few examples of the types of sanity checks that can be used to test an error state extended Kalman filter for navigation using IMU and DVL. It's important to tailor your testing approach to the specific requirements of your application and to ensure that the filter is thoroughly tested before it is deployed in a real-world environment.
Hi Nadav. A good way to check your method and algorithm is to test it on a relatively simple simulation problem, where you know the answers (states) that you are trying to estimate. Add the noise etc in the simulation and check that the filter seems to work right. Once you’re confident it’s working, you can apply it to the real world problems.