Initial commit

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/CovarianceAlgorithmType.cs

@@ -0,0 +1,17 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Covariance estimation algorithm types.
+/// </summary>
+public enum CovarianceAlgorithmType
+{
+    /// <summary>
+    /// Dense SVD algorithm.
+    /// </summary>
+    DenseSvd = 0,
+
+    /// <summary>
+    /// Sparse QR algorithm.
+    /// </summary>
+    SparseQr = 1
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/DenseLinearAlgebraLibraryType.cs

@@ -0,0 +1,22 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Dense linear algebra library types.
+/// </summary>
+public enum DenseLinearAlgebraLibraryType
+{
+    /// <summary>
+    /// Eigen library.
+    /// </summary>
+    Eigen = 0,
+
+    /// <summary>
+    /// LAPACK + BLAS library.
+    /// </summary>
+    Lapack = 1,
+
+    /// <summary>
+    /// NVIDIA's CUDA library.
+    /// </summary>
+    Cuda = 2
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/DoglegType.cs

@@ -0,0 +1,17 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Dogleg type strategies for trust region optimization.
+/// </summary>
+public enum DoglegType
+{
+    /// <summary>
+    /// The traditional approach constructs a dogleg path consisting of two line segments and finds the farthest point on that path that is still within the trust region.
+    /// </summary>
+    TraditionalDogleg = 0,
+
+    /// <summary>
+    /// The subspace approach finds the exact minimum of the model constrained to the subspace spanned by the dogleg path.
+    /// </summary>
+    SubspaceDogleg = 1
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/LineSearchDirectionType.cs

@@ -0,0 +1,27 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Line search direction types for line search minimizer.
+/// </summary>
+public enum LineSearchDirectionType
+{
+    /// <summary>
+    /// Negative of the gradient.
+    /// </summary>
+    SteepestDescent = 0,
+
+    /// <summary>
+    /// A generalization of the Conjugate Gradient method to non-linear functions.
+    /// </summary>
+    NonlinearConjugateGradient = 1,
+
+    /// <summary>
+    /// Limited memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) method.
+    /// </summary>
+    Lbfgs = 2,
+
+    /// <summary>
+    /// Broyden-Fletcher-Goldfarb-Shanno (BFGS) method.
+    /// </summary>
+    Bfgs = 3
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/LineSearchType.cs

@@ -0,0 +1,17 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Line search types for line search minimizer.
+/// </summary>
+public enum LineSearchType
+{
+    /// <summary>
+    /// Backtracking line search with polynomial interpolation or bisection.
+    /// </summary>
+    Armijo = 0,
+
+    /// <summary>
+    /// Wolfe line search.
+    /// </summary>
+    Wolfe = 1
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/LinearSolverType.cs

@@ -0,0 +1,43 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Linear solver types available in Ceres Solver.
+/// Values must match ceres::LinearSolverType enum in Ceres C++ (types.h).
+/// </summary>
+public enum LinearSolverType
+{
+    /// <summary>
+    /// Dense Cholesky factorization of the normal equations.
+    /// </summary>
+    DenseNormalCholesky = 0,
+
+    /// <summary>
+    /// Dense QR factorization.
+    /// </summary>
+    DenseQr = 1,
+
+    /// <summary>
+    /// Sparse normal Cholesky factorization.
+    /// </summary>
+    SparseNormalCholesky = 2,
+
+    /// <summary>
+    /// Dense Schur factorization.
+    /// </summary>
+    DenseSchur = 3,
+
+    /// <summary>
+    /// Sparse Schur factorization.
+    /// </summary>
+    SparseSchur = 4,
+
+    /// <summary>
+    /// Iterative Schur complement solver.
+    /// </summary>
+    IterativeSchur = 5,
+
+    /// <summary>
+    /// Conjugate gradients on the normal equations.
+    /// </summary>
+    CgNr = 6
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/LoggingType.cs

@@ -0,0 +1,17 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Logging types for Ceres Solver output.
+/// </summary>
+public enum LoggingType
+{
+    /// <summary>
+    /// No logging output.
+    /// </summary>
+    Silent = 0,
+
+    /// <summary>
+    /// Log output for each minimizer iteration.
+    /// </summary>
+    PerMinimizerIteration = 1
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/MinimizerType.cs

@@ -0,0 +1,18 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Minimizer types available in Ceres Solver.
+/// Values must match ceres::MinimizerType enum in Ceres C++ (types.h).
+/// </summary>
+public enum MinimizerType
+{
+    /// <summary>
+    /// Line search minimizer.
+    /// </summary>
+    LineSearch = 0,
+
+    /// <summary>
+    /// Trust region minimizer.
+    /// </summary>
+    TrustRegion = 1
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/NonlinearConjugateGradientType.cs

@@ -0,0 +1,22 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Nonlinear conjugate gradient types.
+/// </summary>
+public enum NonlinearConjugateGradientType
+{
+    /// <summary>
+    /// Fletcher-Reeves conjugate gradient.
+    /// </summary>
+    FletcherReeves = 0,
+
+    /// <summary>
+    /// Polak-Ribiere conjugate gradient.
+    /// </summary>
+    PolakRibiere = 1,
+
+    /// <summary>
+    /// Hestenes-Stiefel conjugate gradient.
+    /// </summary>
+    HestenesStiefel = 2
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/NumericDiffMethod.cs

@@ -0,0 +1,26 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Numeric differentiation methods for NumericDiffCostFunction.
+/// Values must match ceres_numeric_diff_method_t enum in ceres_wrapper.h.
+/// Note: The wrapper defines its own enum order (Forward=0, Central=1) which differs
+/// from ceres::NumericDiffMethodType (Central=0, Forward=1). The wrapper uses a switch
+/// statement to map correctly, so C# must match the wrapper's enum.
+/// </summary>
+public enum NumericDiffMethod
+{
+    /// <summary>
+    /// Compute forward finite difference: f'(x) ~ (f(x+h) - f(x)) / h.
+    /// </summary>
+    Forward = 0,
+
+    /// <summary>
+    /// Compute central finite difference: f'(x) ~ (f(x+h) - f(x-h)) / 2h.
+    /// </summary>
+    Central = 1,
+
+    /// <summary>
+    /// Adaptive numerical differentiation using Ridders' method.
+    /// </summary>
+    Ridders = 2
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/PreconditionerType.cs

@@ -0,0 +1,42 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Preconditioner types available in Ceres Solver.
+/// </summary>
+public enum PreconditionerType
+{
+    /// <summary>
+    /// Trivial preconditioner - the identity matrix.
+    /// </summary>
+    Identity = 0,
+
+    /// <summary>
+    /// Block diagonal of the Gauss-Newton Hessian.
+    /// </summary>
+    Jacobi = 1,
+
+    /// <summary>
+    /// Block diagonal of the Schur complement.
+    /// </summary>
+    SchurJacobi = 2,
+
+    /// <summary>
+    /// Power series expansion of the Schur complement.
+    /// </summary>
+    SchurPowerSeriesExpansion = 3,
+
+    /// <summary>
+    /// Visibility clustering based preconditioners.
+    /// </summary>
+    ClusterJacobi = 4,
+
+    /// <summary>
+    /// Cluster tridiagonal preconditioner.
+    /// </summary>
+    ClusterTridiagonal = 5,
+
+    /// <summary>
+    /// Subset preconditioner for general purpose linear least squares problems.
+    /// </summary>
+    Subset = 6
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/SparseLinearAlgebraLibraryType.cs

@@ -0,0 +1,32 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Sparse linear algebra library types.
+/// </summary>
+public enum SparseLinearAlgebraLibraryType
+{
+    /// <summary>
+    /// SuiteSparse library (CHOLMOD).
+    /// </summary>
+    SuiteSparse = 0,
+
+    /// <summary>
+    /// Eigen's sparse linear algebra routines.
+    /// </summary>
+    EigenSparse = 1,
+
+    /// <summary>
+    /// Apple's Accelerate framework sparse linear algebra routines.
+    /// </summary>
+    AccelerateSparse = 2,
+
+    /// <summary>
+    /// NVIDIA's cuDSS and cuSPARSE libraries.
+    /// </summary>
+    CudaSparse = 3,
+
+    /// <summary>
+    /// No sparse linear algebra library should be used.
+    /// </summary>
+    NoSparse = 4
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/TerminationType.cs

@@ -0,0 +1,33 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Termination types for Ceres Solver.
+/// Values must match ceres::TerminationType enum in Ceres C++ (types.h).
+/// </summary>
+public enum TerminationType
+{
+    /// <summary>
+    /// Convergence reached.
+    /// </summary>
+    Convergence = 0,
+
+    /// <summary>
+    /// No convergence.
+    /// </summary>
+    NoConvergence = 1,
+
+    /// <summary>
+    /// Failure.
+    /// </summary>
+    Failure = 2,
+
+    /// <summary>
+    /// User-requested success via IterationCallback returning SOLVER_TERMINATE_SUCCESSFULLY.
+    /// </summary>
+    UserSuccess = 3,
+
+    /// <summary>
+    /// User-requested failure via IterationCallback returning SOLVER_ABORT.
+    /// </summary>
+    UserFailure = 4
+}

srcs/RobotNet10/RobotApp/Communication/CeresSharp/Enums/TrustRegionStrategyType.cs

@@ -0,0 +1,17 @@
+namespace CeresSharp.Enums;
+
+/// <summary>
+/// Trust region strategy types available in Ceres Solver.
+/// </summary>
+public enum TrustRegionStrategyType
+{
+    /// <summary>
+    /// The default trust region strategy is to use the step computation used in the Levenberg-Marquardt algorithm.
+    /// </summary>
+    LevenbergMarquardt = 0,
+
+    /// <summary>
+    /// Powell's dogleg algorithm interpolates between the Cauchy point and the Gauss-Newton step.
+    /// </summary>
+    Dogleg = 1
+}