AIG Software Repository

Artificial Intelligence and Machine Learning Group, UPF, Barcelona

This is the main software repository of the Artificial Intelligence and Machine Learning Research Group at Universitat Pompeu Fabra, Barcelona, Spain. You can browse the full list of publicly available projects on the group’s Github page, or click directly to a particular project on the list below:

Automated Programming Framework

The Automated Programming Framework includes the code necessary to configure and execute different compilations related to the formalisms of Planning Programs and Hierarchical Finite State Controllers. These compilations are produced from the original planning problems using the Universal PDDL Parser, and solved using classical planners which are included in the repository.

Relevant Publications

FS Planner

FS is a classical planner based on the Functional STRIPS planning language, with additional support for existential quantification, state constraints and a fairly large library of global constraints which are useful both from the expressive and computational points of view.

Relevant Publications

The Lightweight Automated Planning Toolkit (LAPKT)

LAPKT is an open-source framework written in C++ / Python to ease the creation of high-performance automated planners. The toolkit has been succesfully used as the basis for a number of planners from the last International Planning Competition (2014), as well as in several research projects on compilation-based and replanning approaches. The full documentation of the toolkit can be found on the project’s main website.

PROBE

PROBE is a classical planner developed by Nir Lipovetzky and Hector Geffner that participated in the seventh International Planning Competition (2011) with remarkable performance. See “Searching for plans with carefully designed probes”, N. Lipovetzky and H. Geffner. In Proceedings ICAPS 2011.

Relevant Publications

Universal PDDL parser

A PDDL parser written in C++ and, used in several of the other software projects within the group. Provides support for programmatically parsing and generating PDDL instances.

Multiagent Parser

Extends the Universal PDDL Parser to provide support for parsing multiagent planning instances expressed in PDDL. Currently also includes a solver for concurrent multiagent planning in which agents can take actions in parallel; the solver is based on a compilation from multiagent planning to classical planning.

Relevant Publications