State initialization code. More...
Classes | |
| class | DynamicInitializer |
| Initializer for a dynamic visual-inertial system. More... | |
| class | Factor_GenericPrior |
| Factor for generic state priors for specific types. More... | |
| class | Factor_ImageReprojCalib |
| Factor of feature bearing observation (raw) with calibration. More... | |
| class | Factor_ImuCPIv1 |
| Factor for IMU continuous preintegration version 1. More... | |
| class | InertialInitializer |
| Initializer for visual-inertial system. More... | |
| struct | InertialInitializerOptions |
| Struct which stores all options needed for state estimation. More... | |
| class | InitializerHelper |
| Has a bunch of helper functions for dynamic initialization (should all be static) More... | |
| class | SimulatorInit |
| Master simulator class that generated visual-inertial measurements. More... | |
| class | State_JPLQuatLocal |
| JPL quaternion CERES state parameterization. More... | |
| class | StaticInitializer |
| Initializer for a static visual-inertial system. More... | |
State initialization code.
Right now this contains StaticInitializer and DynamicInitializer initialization code for a visual-inertial system. It will wait for the platform to stationary, and then initialize its orientation in the gravity frame.
If the platform is not stationary then we leverage dynamic initialization to try to recover the initial state. This is an implementation of the work Estimator initialization in vision-aided inertial navigation with unknown camera-IMU calibration [Dong2012IROS] which solves the initialization problem by first creating a linear system for recovering tthe velocity, gravity, and feature positions. After the initial recovery, a full optimization is performed to allow for covariance recovery. See this tech report for a high level walk through.