Automatic planning is a discipline of artificial intelligence that aims to automatically generate plans, i.e., ordered sequences of actions reaching specific goals. Once generated, the plans can be executed by an agent, a machine, a robot, etc. A generalized plan gives a unique solution to a set of planning problems. In many domains, generalized plans can only make calculations in certain high-level functions, for example, functions that are able to capture key concepts or variables that can accurately distinguish between different possible states. In generalized planning, these high-level functions have to be coded manually. 

A paper presented by Damir Lotinac, Javier Segovia-Aguas, Sergio Jiménez and Anders Jonsson at the International Joint Conference on Artificial Intelligence 2016 (IJCAI 2016) held from 9 to 15 July in New York (USA), contributes a new mechanism that automatically generates high-level functions of state to offer solutions for the problems presented by generalized planning.

This communication is part of a more generic line of work that Anders Jonsson and Hector Geffner (ICREA), principal investigators of the Artificial Intelligence (AI) research group at UPF, are carrying out in the Department of Information and Communication Technologies (DTIC).

A project: Enhancing Usability and Dissemination of Planning Tools, which aims to increase the scope of planning instruments developed in the AI research group, for which, on the one hand this group is working on improving prototypes so that they can be implemented in the productive sector, and on the other, to integrate some of these standard planning tools in the ROS (Robot Operating Systems), the system with which most robots operate today.

Communication of reference:

Damir Lotinac, Javier Segovia-Aguas, Sergio Jiménez and Anders Jonsson “Automatic Generation of High-Level State Features for Generalized Planning”, International Joint Conference on Artificial Intelligence 2016 (IJCAI 2016) 9-15 July, New York (USA).