precalculateParams.m

function precalculatedParams = precalculateParams(scenarioParam, filterParam)
% @author Florian Pfaff pfaff@kit.edu
% @date 2016-2023
% V3.1
arguments (Input)
    scenarioParam (1,1) struct
    filterParam (1,1) struct
end
arguments (Output)
    precalculatedParams struct 
end
precalculatedParams = struct();
switch filterParam.name
    case 'twn'
        if ~scenarioParam.useLikelihood && ~isa(scenarioParam.measNoise, 'ToroidalWNDistribution')
            s = scenarioParam.measNoise.sample(100000);
            precalculatedParams.measNoiseForFilter = ToroidalWNDistribution.mleNumerical(s);
        end
    case 'htgf'
        if scenarioParam.useTransition
            precalculatedParams.fTrans_tdxtd = ...
                TdCondTdGridDistribution.fromFunction(scenarioParam.fTrans, filterParam.parameter, true, 'CartesianProduct', 2*scenarioParam.initialPrior.dim);
        end
    case 'iff'
        filtertmp = HypertoroidalFourierFilter(repmat(filterParam.parameter, [1, scenarioParam.initialPrior.dim]), 'identity');
        if scenarioParam.useTransition
            precalculatedParams.hfdTrans = ...
                filtertmp.getfTransAsHfd(scenarioParam.genNextStateWithoutNoiseFourier, scenarioParam.sysNoise);
        end
    case 'sqff'
        filtertmp = HypertoroidalFourierFilter(repmat(filterParam.parameter, [1, scenarioParam.initialPrior.dim]), 'sqrt');
        if scenarioParam.useTransition
            precalculatedParams.hfdTrans = ...
                filtertmp.getfTransAsHfd(scenarioParam.genNextStateWithoutNoiseFourier, scenarioParam.sysNoise);
        end
    case 'sgf'
        if scenarioParam.useTransition
            precalculatedParams.fTrans_sdxsd = ...
                S2CondS2GridDistribution.fromFunction(scenarioParam.fTrans, filterParam.parameter, true, 'eq_point_set');
        end
    case 'hgf'
        if scenarioParam.useTransition
            precalculatedParams.fTrans_sdxsd = ...
                SdCondSdGridDistribution.fromFunction(scenarioParam.fTrans, filterParam.parameter, true, 'eq_point_set', 2*scenarioParam.initialPrior.dim);
        end
    case 'hgfSymm'
        if scenarioParam.useTransition
            precalculatedParams.fTrans_sdxsd = ...
                SdCondSdGridDistribution.fromFunction(scenarioParam.fTrans, filterParam.parameter, true, 'eq_point_set_symm', 2*scenarioParam.initialPrior.dim);
        end
    case 'hhgf'
        precalculatedParams.priorForFilter = HyperhemisphericalGridDistribution.fromDistribution(scenarioParam.initialPrior, filterParam.parameter, 'eq_point_set_symm');
        if scenarioParam.useTransition
            precalculatedParams.fTrans_sdhalfxsdhalf = ...
                SdHalfCondSdHalfGridDistribution.fromFunction( ...
                scenarioParam.fTrans, filterParam.parameter, true, 'eq_point_set_symm', 2*scenarioParam.initialPrior.dim);
        end
    case 's3f'
        switch scenarioParam.manifoldType
            case 'se2'
                precalculatedParams.priorForFilter = SE2StateSpaceSubdivisionGaussianDistribution( ...
                    FIGDistribution.fromDistribution(scenarioParam.initialPriorPeriodic, filterParam.parameter), ...
                    repmat(scenarioParam.initialPriorLinear, [filterParam.parameter, 1]));
                precalculatedParams.condPeriodic = ...
                    TdCondTdGridDistribution.fromFunction(@(xkk, xk)scenarioParam.sysNoisePeriodic.pdf(xkk - xk), filterParam.parameter, false, 'CartesianProd', 2);
            case 'se3'
                precalculatedParams.priorForFilter = SE3StateSpaceSubdivisionGaussianDistribution( ...
                    HyperhemisphericalGridDistribution.fromDistribution(scenarioParam.initialPriorPeriodic, filterParam.parameter), ...
                    repmat(scenarioParam.initialPriorLinear, [filterParam.parameter, 1]));
                precalculatedParams.condPeriodic = ...
                    SdHalfCondSdHalfGridDistribution.fromFunction(@(xkk, xk)cell2mat(arrayfun(@(i){scenarioParam.sysNoisePeriodic.shift(xk(:,i)).pdf(xkk)},1:size(xk,2))), filterParam.parameter, true, 'eq_point_set_symm', 8);
            otherwise
                error('Manifold type not supported.')
        end
    case 'se2bf'
        noSamplesForFitting = 100000;
        precalculatedParams.priorForFilter = SE2BinghamDistribution.fit(scenarioParam.initialPrior.sample(noSamplesForFitting));
        precalculatedParams.sysNoiseForFilter = SE2BinghamDistribution.fit(scenarioParam.sysNoise.sample(noSamplesForFitting));
end
end