Statistical Filter
Validation logic applied to incoming sensor data prevents the assimilation of outlier measurements into a state estimation filter. Applying chi-squared gating ensures that only observations with a high probability of being correct are processed. The procedure calculates a normalized distance between the predicted measurement and the actual observation.
Rejection Threshold
A specific numerical bound defines whether a measurement is accepted or discarded. This threshold is derived from the chi-squared distribution table based on the degrees of freedom in the measurement vector. When the calculated residual exceeds this value, chi-squared gating identifies the data point as a spurious signal.
Operational Mechanism
The process begins with the calculation of the innovation vector, which represents the difference between the expected and actual sensor reading. This innovation is then scaled by the inverse of the innovation covariance matrix to produce a scalar value. Chi-squared gating uses this scalar to assess the likelihood of the measurement being part of the expected noise distribution.
If the value is too high, the filter ignores the update and relies on its internal model. This prevents sensor glitches or multipath errors from corrupting the navigation solution.
System Reliability
Resilience in autonomous systems depends on the ability to ignore faulty inputs. Proper tuning of chi-squared gating balances the risk of rejecting valid data against the risk of accepting false positives.
