Revision 65 as of 2019-06-04 16:41:05

Clear message

Back to the HomePage.

Obtaining and running Fast Downward

This page describes how to build and run the Fast Downward planner.

Supported platforms

The planner is mainly developed under Linux; and all of its features should work with no restrictions under this platform. The planner should compile and run correctly on macOS, but we cannot guarantee that it works as well as under Linux. The same comment applies for Windows, where additionally some diagnostic features (e.g. reporting memory used when the planner is terminated by a signal) are not supported. Setting time and memory limits and running portfolios is not supported under Windows either.

We appreciate bug reports and patches for all platforms, in particular, contributions to the documentation (e.g. installation instructions) for macOS and Windows.

Getting access to the planner

Since October 2010, the main development of the code takes place in a Mercurial repository. Before then, the code was hosted in a Subversion repository, which is still available. However, the trunk of the Subversion repository has been frozen (set to read-only) and will no longer receive updates or bug fixes. See LegacySubversionRepository if you need to access the old repository for some reason.

If you haven't used Mercurial before, you may want to have a look at the excellent documentation available online (also available as a printed book published by O'Reilly Media). However, you don't really have to understand Mercurial to use the planner. For those experienced with Mercurial or similar distributed version control systems, we currently use a "single-pusher" task-based workflow with a single pusher for the Fast Downward repository. If you want to contribute to Fast Downward, the recommended way is to set up a clone of the master repository (e.g. on Bitbucket) and provide a link to it in our issue tracker so that we can pull from it.

Dependencies

Linux: To obtain and build the planner, you will need the Mercurial version control system, a C++11 compiler, CMake and GNU make. To run the planner, you will also need Python 2.7 or Python >= 3.2. On Linux, the following should install all these dependencies: 

sudo apt install cmake g++ mercurial make python

Windows: If using Windows, you should install Visual Studio >= 2017, Python, Mercurial, and CMake. During the installation of Visual Studio, the C++ compiler is not installed by default, but the IDE will prompt you to install it when you create a new C++ project.

Linear-Programming configurations

Some configurations require an LP solver to work. The planner will compile fine if there is no LP installed on the system, but trying to use the features that require an LP solver will generate an error message explaining what is missing. See LPBuildInstructions for instructions on how to set up an LP solver and tell Fast Downward about it.

Validating the computed plans

You can validate the found plans by passing --validate to the planner. For this, the VAL plan validation software needs to be installed on your system. Here you can find some instructions to help you set it up.

Obtaining the code

The command 

hg clone http://hg.fast-downward.org DIRNAME

will create a clone of the Fast Downward master repository in directory DIRNAME. The directory is created if it does not yet exist. In the following, we assume that you used downward as the DIRNAME.

Compiling the planner

To build the planner for the first time, run: 

   1 cd downward
   2 ./build.py

This will create our default build release in the directory downward/builds. Other predefined build types are debug, releasenolp, debugnolp and minimal. Calling ./build.py --debug will create a default debug build (equivalent to debug). You can pass make parameters to ./build.py, e.g., ./build.py --debug -j4 will create a debug build using 4 threads for compilation (-j4), and ./build.py translate will only build the translator component. (Because the translator is implemented in Python, "building" it just entails copying its source code into the build directory.) By default, build.py uses all cores for building the planner.

See #Manual_Builds for more complex builds.

Compiling on macOS

Fast Downward should compile with the GNU C++ compiler and clang with the same instructions described above. In case your system compiler does not work for some reason, you will need to install the GNU compiler (if not already present) and tell CMake which compiler to use (see #ManualBuilds).

Compiling on Windows

Windows does not interpret the shebang in python files, so you have to call build.py as python build.py (make sure python is on your PATH). Also note that options are passed without --, e.g., python build.py build=releasenolp. The steps above should work if you execute them in the developer console of Visual Studio which will use nmake and cl for the build. For Visual Studio 2017 you have to open a regular shell (cmd) and call the script C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvarsall.bat x64 to set up the build environment.

Alternatively, you can create a Visual Studio Project (see #Manual_Builds), open it in Visual Studio and build from there. Visual Studio will create its binary files in subdirectories of the project that our driver script currently does not recognize. If you build with Visual Studio, you will have to run the individual components of the planner yourself.

Manual and Custom Builds

The build.py script only creates a directory and calls cmake and make. To do these steps manually, run: 

   1 mkdir -p builds/mycustombuild
   2 cd builds/mycustombuild
   3 cmake ../../src CMAKE_OPTIONS
   4 make

where CMAKE_OPTIONS are the options used to configure the build (see below). Without options, this results in the release build. (Use --build mycustombuild in the fast-downward.py script to select this build when running the planner.)

You can use a CMake GUI to set up all available options. To do so, on Unix-like systems replace the call to cmake by ccmake (sudo apt install cmake-curses-gui). On Windows, open the CMake GUI and enter the paths there.

Possible options to configure the build include:

You can also generate makefiles for other build systems (such as ninja) or generate project files for most IDEs:

You can also change the compiler/linker and their options by setting the environment variables CC, CXX, LINKER and CXXFLAGS. These variables need to be set before running ./build.py or executing cmake manually. If you want to change these settings for an existing build, you must manually remove the build directory before rerunning ./build.py.

Examples:

If you use a configuration often, it might make sense to add an alias for it in build_configs.py.

Running the planner

For basic instructions on how to run the planner including examples, see PlannerUsage. The search component of the planner accepts a host of different options with widely differing behaviour. At the very least, you will want to choose a search engine with one or more evaluator specifications.

Caveats

Please be aware of the following issues when working with the planner, especially if you want to use it for conducting scientific experiments:

  1. You are working with the most recent version of the development branch of the code, which is usually not thoroughly tested. Things can break or degrade with every commit. Typically they don't, but if they do, don't be surprised. We are working on a proper release, but we are not there yet. For more information, see the list of outstanding issues for release 1.0 of the planner.

  2. There are known bugs, especially with the translator component. To find out more, check out our issue tracker. The planner has only really been tested with IPC domains, and even for those it does not work properly with all formulations of all domains. For more information, see the section on #Known_good_domains below.

  3. The search options are built with flexibility in mind, not ease of use. It is very easy to use option settings that look plausible, yet introduce significant inefficiencies. For example, an invocation like 

    ./fast-downward.py domain.pddl problem.pddl --search "lazy_greedy([ff()], preferred=[ff()])"

    looks plausible, yet is hugely inefficient since it will compute the FF heuristic twice per state. See the examples on the PlannerUsage page to see how to call the planner properly. If in doubt, ask.

Known good domains

There is a large collection of planning competition benchmarks at https://bitbucket.org/aibasel/downward-benchmarks, which includes all IPC domains (but not all formulations of all domains). The planner is somewhat sensitive to non-STRIPS language constructs and will choke on some valid PDDL inputs. Moreover, many of the heuristics do not support axioms or conditional effects. Even worse, sometimes the translator will introduce conditional effects even though the original PDDL input did not contain them. (We are working on this.)

We recommend that you use the sets of domains that are predefined in the suites.py script for your experiments. Here are the benchmark suites that we usually use in the two most common settings:

After cloning the repo, you can list the domains in the respective suites by doing: 

   1 ./suites.py optimal_strips

You can use the resulting list of domains in your experiment scripts (see below).

Experiments with Fast Downward

The Downward Lab toolkit helps running Fast Downward experiments on large benchmark sets. → More information