Meaning
An estimation algorithm provides simultaneous real-time tracking of both dynamic system variables and slowly varying internal parameters from noisy sensor measurements. The dual state kalman observer achieves this by maintaining two separate filter processes that run in parallel or in a cascaded loop to update the system state vector alongside the parameter vector. This algorithm establishes a structured approach for handling non-linearities and parameter variations in connected drive modules.
It operates up to the frequency limit where processor cycle times can no longer execute the parallel matrix multiplications.
Estimation Architecture
Separate estimation loops decouple the fast dynamic states, such as velocity or current, from the slow parameter changes like thermal resistance or winding degradation. This dual state kalman observer structure avoids the high dimensional matrix operations typical of a single fully extended filter by partition of the state space. Computational execution becomes highly deterministic, ensuring predictable cycle times on embedded microcontrollers.
The partitioned filter structure runs efficiently on standard ARM Cortex processors.
Mathematical Convergence
Proper execution of the algorithm depends on the distinct timescales of the state and parameter changes. A system designer configures the parameter update loop to run at a lower frequency or with a higher covariance than the state update loop. This distinction prevents the parameter updates from introducing artificial instability into the high speed control loops of the module.
When timescales are correctly separated, the system converges reliably to stable estimates.
Hardware Integration
High speed data acquisition must provide synchronised samples of current and position to the observer routine to prevent estimation drift. The dual state kalman observer uses these hardware interrupts to execute its correction steps before the next duty cycle update. Verification involves testing the algorithm against known hardware faults or temperature steps on a dynamometer.
This process ensures the software model matches the physical hardware under extreme loads. During integration testing, the developer monitors the execution overhead to guarantee that the observer routine completes within its allocated microprocessor slot. If the execution budget is exceeded, the controller may fail to update the pulse-width modulation signals, leading to audible motor noise or severe control instability.