plevasseur / rudie2.0 Goto Github PK

Blockers

• Improving the docs
• Adding CI/CD pipeline

Main benefit of the library is due to the ability to build custom filters from swappable library of models

Process Models

• ConstantVelocityModel2d
• ConstantTurnRateVelocityModel2d

Measurement Models

• RadarMeasurementModel2d
• GpsMeasurementModel2d

• Investigate what's used these days and is free
• Set it up

Priority list

• Improve docs 'splash page' with a simple example as a doctest
• Simplest place to start -- struct_based ExtendedKalmanFilter
• How to write a new process model
• How to write a new measurement model
• How to use the generate_nonlinear_predict_chain_custom proc macro
• (When it's written) How to use the generate_nonlinear_update_sequential_custom

Would work akin to the similar proc macro for generate_nonlinear_predict_chain_custom for a library user

Open issues

• How to handle modifying non-contiguous memory of the state vector and state covariance (VectorViewMut and MatrixViewMut appear to only be able to handle snagging contiguous memory and an update step may need to access sparse sections of the state and covariance)
• Proc macro needs to generate a custom trait with multiple default impl'ed predict functions, 1 x measurement (complicated)

Implement machinery needed for Unscented Kalman Filter

Align the NonlinearUpdate trait in struct_based in a similar manner to how NonlinearPredict is now aligned with custom proc macro approach

Blocked on

• Completion of NonlinearUpdate custom proc macro