[Manifolds](http://en.wikipedia.org/wiki/Manifold#Charts.2C_atlases.2C_and_transition_maps) and [charts](http://en.wikipedia.org/wiki/Manifold#Charts.2C_atlases.2C_and_transition_maps) are intimately linked concepts. We are only interested here in [differentiable manifolds](http://en.wikipedia.org/wiki/Differentiable_manifold#Definition), continuous spaces that can be locally approximated *at any point* using a local vector space, called the [tangent space](http://en.wikipedia.org/wiki/Tangent_space). A *chart* is an invertible map from the manifold to that tangent space.
In GTSAM, all properties and operations needed to use a type must be defined through template specialization of the struct `gtsam::traits`. Concept checks are used to check that all required functions are implemented.
*`enum { dimension = D};`, an enum that indicates the dimensionality *n* of the manifold. In Eigen-fashion, we also support manifolds whose dimenionality is only defined at runtime, by specifying the value -1.
*`ChartJacobian`, a typedef for `OptionalJacobian<dimension, dimension>`.
*`ManifoldType`, a pointer back to the type.
*`structure_category`, a tag type that defines what requirements the type fulfills, and therefore what requirements this traits class must fulfill. It should be defined to be one of the following:
*`gtsam::traits::manifold_tag` -- Everything in this list is expected
*`v = traits<T>::Local(p,q)`, the chart, from manifold to tangent space, think of it as *q (-) p*, where *p* and *q* are elements of the manifold and the result, *v* is an element of the vector space.
*`p = traits<T>::Retract(p,v)`, the inverse chart, from tangent space to manifold, think of it as *p (+) v*, where *p* is an element of the manifold and the result, *v* is an element of the vector space.
A [group]("http://en.wikipedia.org/wiki/Group_(mathematics)"") should be well known from grade school :-), and provides a type with a composition operation that is closed, associative, has an identity element, and an inverse for each element. The following should be added to the traits class for a group:
We do *not* at this time support more than one composition operator per type. Although mathematically possible, it is hardly ever needed, and the machinery to support it would be burdensome and counter-intuitive.
Also, a type should provide either multiplication or addition operators depending on the flavor of the operation. To distinguish between the two, we will use a tag (see below).
For Lie groups, the `exponential map` above is the most obvious mapping: it
associates straight lines in the tangent space with geodesics on the manifold
(and it's inverse, the log map). However, there are two cases in which we deviate from this:
However, the exponential map is unnecessarily expensive for use in optimization. Hence, in GTSAM there is the option to provide a cheaper chart by means of the `ChartAtOrigin` struct in a class. This is done for *SE(2)*, *SO(3)* and *SE(3)* (see `Pose2`, `Rot3`, `Pose3`)
Most Lie groups we care about are *Matrix groups*, continuous sub-groups of *GL(n)*, the group of *n x n* invertible matrices. In this case, a lot of the derivatives calculations needed can be standardized, and this is done by the `LieGroup` superclass. You only need to provide an `AdjointMap` method.
While vector spaces are in principle also manifolds, it is overkill to think about charts etc. Really, we should simply think about vector addition and subtraction. I.e.where
This considerably simplifies certain operations. A `VectorSpace` superclass is available to implement the traits. Types that are vector space models include `Matrix`, `Vector`, any fixed or dynamic Eigen Matrix, `Point2`, and `Point3`.
