We provide here implementations of RRG [1] and the optimal extension [2] for motion planning subject to deterministic μ-Calculus constraints. The latter paper makes use of important results from [3]. The implementation includes corrections to and missing details from [2].
-
S. Karaman and E. Frazzoli (2009). Sampling-based Motion Planning with Deterministic μ-Calculus Specifications. in Proceedings of the 48th IEEE Conference on Decision and Control (CDC), pp. 2222-2229. DOI: 10.1109/CDC.2009.5400278
-
S. Karaman and E. Frazzoli (2012). Sampling-based Algorithms for Optimal Motion Planning with Deterministic μ-Calculus Specifications. in Proceedings of the American Control Conference (ACC), pp. 735-74. DOI: 10.1109/ACC.2012.6315419
-
S. Karaman and E. Frazzoli (2011). Sampling-based Algorithms for Optimal Motion Planning. International Journal of Robotics Research 30(7): 846-894. DOI: 10.1177/0278364911406761
The usual CMake idiom,
mkdir build
cd build
cmake ..
On a standard UNIX platform, you can next run make to build all examples, the
programs for which will be under bin/. To instead build a particular example,
call make with its name, e.g., to build only the rrg_dubins_car example,
make rrg_dubins_car
To build the API manual,
make doc
which requires Doxygen to be installed. The generated files will be under the directory html/
Current results of continuous integration (CI) testing:
Thanks to Travis CI and AppVeyor for CI services.
The rrglib library consists primarily of header files and does not require
compilation a priori. However, support is provided for installing header files
and libraries in a common location and against which other programs can
link. The default CMake install prefix can be changed by providing a value for
CMAKE_INSTALL_PREFIX. E.g., cmake -DCMAKE_INSTALL_PREFIX=~/opt .. will use the
opt/ directory in your home directory instead of the system-wide /usr or
/usr/local. HPP files will be placed under include/rrglib/ (after the install
prefix), and compiled object and CMake configuration files will be installed
under lib/rrglib/
For example, to use it in a CMakeLists.txt
find_package (rrglib REQUIRED)
include_directories (~/opt/include/rrglib)
add_executable (rrgstar_double_integrator rrgstar_doubleinteg.cpp)
target_link_libraries (rrgstar_double_integrator kdtree incmumc)
where incmumc is part of rrglib. kdtree is not a part of rrglib (it originates elsewhere) but is shipped with rrglib for convenience.
Note that visualization is not intended to be the main purpose of this repository, so Mayavi, Matplotlib, etc. as used in tools/rrgplot.py are regarded as optional dependencies. Similarly, building the various interfaces may require external resources to be available, like ROS. However, those dependencies are only necessary if you want to build that particular interface.
The primary means for visualizing is tools/rrgplot.py. For 2D plotting it uses
Matplotlib, and for 3D plotting
Mayavi. The former is part of the
standard SciPy stack.
After completing the installation instructions above, you can run examples and plot the resulting trajectories. To use the RRG planner for Dubins car, from the build/ directory, try
bin/rrg_dubins_car > tmp.json
../tools/rrgplot.py ../examples/2d3r.json tmp.json
To use the RRG* planner for the same dynamics model (Dubins car),
bin/rrgstar_dubins_car > tmp.json
../tools/rrgplot.py ../examples/2d3r.json tmp.json
Scott C. Livingston (Caltech),
and Sertac Karaman and Emilio Frazzoli (MIT), who wrote the code on which the present work is based and as available from https://svn.csail.mit.edu/smp, which has last changed date of 2011-08-02 05:25:23 -0400. The rrgstar repository began by importing from there.
This is free software released under the terms of the MIT License. There is no warranty; not even for merchantability or fitness for a particular purpose. Consult LICENSE for copying conditions.
Copies of several third-party free, open-source softwares are included under the directory src/external_libraries/. References to original locations and licenses can be found there.