Java generics are too limited to do what we need. This refactors generic code previously in Unit and Measure into unit-specific classes that can have unit-safe math operations (notably, times and divide) that can return values in known units instead of a wildcarded Measure<?>.
Unit-specific measure implementations are automatically generated by ./wpiunits/generate_units.py, which generates generic interfaces and mutable and immutable implementations of those interfaces. These make up the bulk of the diff of this PR (approximately 9300 LOC).
This also adds units for angular and linear velocities, accelerations, and momenta; moment of inertia; and torque.
This makes it easier to define schemas when the type name is non-trivial (e.g., templated structs).
This is breaking for a) custom struct implementations and b) anything calling `wpi::Struct<T>::GetTypeString(info...)` in C++ directly. In both cases, it's a simple translation: For A, rename `GetTypeString()` to `GetTypeName()` and remove the struct: at the beginning, and for B, use `wpi::GetStructTypeString<T>(info...)` instead.
C++ doesn't need this because it supports value types, which are much
cheaper to construct. constexpr is also available to make construction
zero-cost.
LTVUnicycleController is a drop-in replacement with better tuning knobs.
The RamseteCommand examples were removed instead of retrofitted with
LTVUnicycleController because we're planning on removing the command
controller classes anyway, so it would be wasted effort. The
SimpleDifferentialDriveSimulation example shows direct
LTVUnicycleController usage.
This implements de/serialization for the types that aren't templated (SwerveDriveKinematics) in C++ or where there is no trivial way to go round-trip (Splines) for the messages.
ProfiledPIDController and ExponentialProfile use current, then goal.
This isn't a breaking change because this overload of calculate() is
new for 2024.
Made JNI modifications to expose the faster function, made the API use
the typesafe Matrix API, and synchronized the documentation with C++.
Sped up C++ LTV diff drive test from 20 ms to 15 ms.
Sped up C++ LTV unicycle test from 15 ms to 10 ms.
15 m/s is about 50 ft/s, which is way above what FRC robots should be
able to achieve. This limit lets us catch user errors from bad unit
conversions immediately instead of the LUT generation in the LTV
controllers hanging for a really long time.
Fixes#5027.
I timed the DARE unit tests, and the new solver is 0 to 100% faster in
all cases (that is, it's at least as fast as Drake's and up to 2x faster
in some cases).
The new solver is also much simpler, takes less time to compile, and
drops the libwpimath.so size from 325 MB to 301 MB.
I think most of the compilation time is coming from the eigenvalue
decompositions used to enforce argument preconditions.
