Added two ReadMe files to document the plethora of smart factors.

gtsam/slam/ReadMe.md

@@ -0,0 +1,64 @@
+# SLAM Factors
+
+## GenericProjectionFactor (defined in ProjectionFactor.h)
+
+Non-linear factor for a constraint derived from a 2D measurement. 
+The calibration is assumed known and passed in the constructor.
+The main building block for visual SLAM.
+
+Templated on
+- `POSE`, default `Pose3`
+- `LANDMARK`, default `Point3`
+- `CALIBRATION`, default `Cal3_S2`
+
+## SmartFactors
+
+These are "structure-less" factors, i.e., rather than introducing a new variable for an observed 3D point or landmark, a single factor is created that provides a multi-view constraint on several poses and/or cameras.
+
+### SmartFactorBase
+
+This is the base class for smart factors, templated on a `CAMERA` type.
+It has no internal point, but it saves the measurements, keeps a noise model, and an optional sensor pose.
+
+### SmartProjectionFactor
+
+Also templated on `CAMERA`. Triangulates a 3D point and keeps an estimate of it around.
+This factor operates with monocular cameras, and is used to optimize the camera pose
+*and* calibration for each camera, and requires variables of type `CAMERA` in values.
+
+If the calibration is fixed use `SmartProjectionPoseFactor` instead!
+
+
+### SmartProjectionPoseFactor
+
+Derives from `SmartProjectionFactor` but is templated on a `CALIBRATION` type, setting `CAMERA = PinholePose<CALIBRATION>`.
+This factor assumes that the camera calibration is fixed and the same for all cameras involved in this factor.
+The factor only constrains poses.
+
+If the calibration should be optimized, as well, use `SmartProjectionFactor` instead!
+
+### SmartProjectionFactorP
+
+Same as `SmartProjectionPoseFactor`, except:
+- it is templated on `CAMERA`, i.e., it allows cameras beyond pinhole;
+- it admits a different calibration for each measurement, i.e., it can model a multi-camera system;
+- it allows multiple observations from the same pose/key, again, to model a multi-camera system.
+
+TODO: DimPose and ZDim are hardcoded. Copy/paste from `SmartProjectionPoseFactor`. Unclear what the use case is.
+
+### RegularImplicitSchurFactor
+
+A specialization of a GaussianFactor to structure-less SFM, which is very fast in a conjugate gradient (CG) solver. 
+It is produced by calling `createRegularImplicitSchurFactor` in `SmartFactorBase` or `SmartProjectionFactor`. 
+
+### JacobianFactorQ
+
+A RegularJacobianFactor that uses some badly documented reduction on the Jacobians.
+
+### JacobianFactorQR
+
+A RegularJacobianFactor that eliminates a point using sequential elimination.
+
+### JacobianFactorQR
+
+A RegularJacobianFactor that uses the "Nullspace Trick" by Mourikis et al. See the documentation in the file, which *is* well documented.

gtsam_unstable/slam/ReadMe.md

@@ -0,0 +1,39 @@
+# SLAM Factors
+
+## SmartFactors
+
+These are "structure-less" factors, i.e., rather than introducing a new variable for an observed 3D point or landmark, a single factor is created that provides a multi-view constraint on several poses and/or cameras.
+
+### SmartRangeFactor
+
+An experiment in creating a structure-less 2D range-SLAM factor with range-only measurements.
+It uses a sophisticated `triangulate` logic based on circle intersections.
+
+### SmartStereoProjectionFactor
+
+Version of `SmartProjectionFactor` for stereo observations, specializes SmartFactorBase for `CAMERA == StereoCamera`.
+
+TODO: a lot of commented out code and could move a lot to .cpp file.
+
+### SmartStereoProjectionPoseFactor
+
+Derives from `SmartStereoProjectionFactor` but adds an array of `Cal3_S2Stereo` calibration objects .
+
+TODO: Again, as no template arguments, we could move a lot to .cpp file.
+
+### SmartStereoProjectionFactorPP
+
+Similar `SmartStereoProjectionPoseFactor` but *additionally* adds an array of body_P_cam poses. The dimensions seem to be hardcoded and the types defined in the SmartFactorBase have been re-defined.  
+
+TODO: See above, same issues as `SmartStereoProjectionPoseFactor`.
+
+### SmartProjectionPoseFactorRollingShutter
+
+Is templated on a `CAMERA` type and derives from `SmartProjectionFactor`.
+
+This factor optimizes two consecutive poses of a body assuming a rolling
+shutter model of the camera with given readout time. The factor requires that
+values contain (for each 2D observation) two consecutive camera poses from
+which the 2D observation pose can be interpolated.
+
+TODO: the dimensions seem to be hardcoded and the types defined in the SmartFactorBase have been re-defined. Also, possibly a lot of copy/paste computation of things that (should) happen in base class.