Member LegacyDataPacket::m_triggerIn1

remove this

Member LegacyDataPacket::m_triggerIn2

remove this

Add the internal buffer to the Windows implementation also

Class mrpt::config::CLoadableOptions

Automatize this class thru a proxy auxiliary class where variables are registered from pointers, etc...

Member mrpt::graphslam::CGraphSlamEngine< GRAPH_T >::setCurrentPositionModel (const std::string &model_name, const mrpt::img::TColor &model_color=mrpt::img::TColor(0, 0, 0), const size_t model_size=1, const pose_t &init_pose=pose_t())

Use an airplane/quad model for 3D operations

Member mrpt::graphslam::deciders::CLoopCloserERD< GRAPH_T >::updateMapPartitionsVisualization ()

Have some sorts of a string_view instead

Member mrpt::maps::CLandmarksMap::TCustomSequenceLandmarks::m_grid

Use the KD-tree instead?

Member mrpt::maps::CMultiMetricMapPDF::prediction_and_update_pfOptimalProposal (const mrpt::obs::CActionCollection *action, const mrpt::obs::CSensoryFrame *observation, const bayes::CParticleFilter::TParticleFilterOptions &PF_options) override

Add paper ref!

If there are 2+ sensors on the robot, compute phi?

Class mrpt::nav::PlannerRRT_SE2_TPS

Factorize into more generic path planner classes! //template <class POSE, class MOTIONS>...

Member mrpt::opengl::CPolyhedron::CreateCatalanDoublePyramid (uint32_t numBaseEdges, double height)

Actually resulting height is significantly higher than that passed to the algorithm.

Member mrpt::opengl::CPolyhedron::CreateCatalanTrapezohedron (uint32_t numBaseEdges, double height)

Actually resulting height is significantly higher than that passed to the algorithm.

Member PacketStamper::stampPacket (XsDataPacket &pack, XsDataPacket &highest)

This could be a (couple of) milliseconds too late, this should be set as soon as the source message arrives: mantis 7157

Member ProtocolHandler::composeMessage (XsByteArray &raw, const XsMessage &msg)

Generalize this method -> IProtocolHandler

Member ProtocolHandler::findMessage (XsMessage &rcv, const XsByteArray &raw) const

Since the assumption is that we receive a stream of valid messages without garbage, the scan is implemented in a rather naive and simple way. If we can expect lots of garbage in the data stream, this should probably be looked into.