-
Notifications
You must be signed in to change notification settings - Fork 10
/
USAGE
81 lines (63 loc) · 3.61 KB
/
USAGE
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
USAGE - Georgia Tech Smoothing and Mapping library
---------------------------------------------------
What is this file?
This file explains how to make use of the library for common SLAM tasks,
using a visual SLAM implementation as an example.
Getting Started
---------------------------------------------------
Install:
Follow the installation instructions in the README file to build and
install gtsam, as well as running tests to ensure the library is working
properly.
Compiling/Linking with gtsam:
The installation creates a binary "libgtsam" at the installation prefix,
and an include folder "gtsam". These are the only required includes, but
the library has also been designed to make use of XML serialization through
the Boost.serialization library, which requires the the Boost.serialization
headers and binaries to be linked.
If you use CMake for your project, you can use the CMake scripts in the
cmake folder for finding GTSAM, CppUnitLite, and Wrap.
Examples:
To see how the library works, examine the unit tests provided.
Overview
---------------------------------------------------
The GTSAM library has three primary components necessary for the construction
of factor graph representation and optimization which users will need to
adapt to their particular problem.
FactorGraph:
A factor graph contains a set of variables to solve for (i.e., robot poses,
landmark poses, etc.) and a set of constraints between these variables, which
make up factors.
Values:
Values is a single object containing labeled values for all of the
variables. Currently, all variables are labeled with strings, but the type
or organization of the variables can change
Factors:
A nonlinear factor expresses a constraint between variables, which in the
SLAM example, is a measurement such as a visual reading on a landmark or
odometry.
The library is organized according to the following directory structure:
3rdparty local copies of third party libraries - Eigen3 and CCOLAMD
base provides some base Math and data structures, as well as test-related utilities
geometry points, poses, tensors, etc
inference core graphical model inference such as factor graphs, junction trees, Bayes nets, Bayes trees
linear inference specialized to Gaussian linear case, GaussianFactorGraph etc...
nonlinear non-linear factor graphs and non-linear optimization
slam SLAM and visual SLAM application code
VSLAM Example
---------------------------------------------------
The visual slam example shows a full implementation of a slam system. The example contains
derived versions of NonlinearFactor, NonlinearFactorGraph, in classes visualSLAM::ProjectionFactor,
visualSLAM::Graph, respectively. The values for the system are stored in the generic
Values structure. For definitions and interface, see gtsam/slam/visualSLAM.h.
The clearest example of the use of the graph to find a solution is in
testVSLAM. The basic process for using graphs is as follows (and can be seen in
the test):
- Create a NonlinearFactorGraph object (visualSLAM::Graph)
- Add factors to the graph (note the use of Boost.shared_ptr here) (visualSLAM::ProjectionFactor)
- Create an initial configuration (Values)
- Create an elimination ordering of variables (this must include all variables)
- Create and initialize a NonlinearOptimizer object (Note that this is a generic
algorithm that does not need to be derived for a particular problem)
- Call optimization functions with the optimizer to optimize the graph
- Extract an updated values from the optimizer