César Guzmán2, Vidal Alcázar1, David Prior3, Eva Onaindía2, Daniel Borrajo1, Juan Fdez-Olivares3.
Universidad Carlos III de Madrid1, Universidad Politècnica de València2, Universidad de Granada3
This paper appears in: 21th International Conference on Automated Planning and Scheduling, ICAPS - Systems Demo. Freiburg, Germany.
Issue Date: Jun.11-16, 2011
Abstract - download article
In this work, we present our ongoing effort on building a domain-independent software platform that integrates basic capabilities for planning, execution, monitoring, re-planning and learning. We name it PELEA after Planning, Execution and LEarning Architecture. The goal is two-fold: first, to provide software engineers a tool that can be used off-the-shelf to easily build planning applications, supporting a rapid prototyping life-cycle; and second to provide planning practitioners a tool that can be highly configured and in which new components replacing the ones that are already integrated can be easily added. Regarding the first goal, the platform currently includes state-of-the-art components for performing a wide range of (meta-)planning tasks, such as: planning (using several paradigms), controlled execution, monitoring of correct plan execution, re-planning when needed, learning of control knowledge, or low-level planning. Ultimately the user could use the tool as-is by giving as input a domain and problem descriptions. Regarding the second goal, it can serve as a benchmark platform for comparing different techniques under the same conditions. For example, a planning expert might want to try out a new re-planning technique on a robot simulator without the need to generate a complete planning-execution-monitoring-replanning architecture. We are currently interfacing the platform with known simulators (videogames and robotic platforms) as well as developing new ones for specific domains (logistics) and even a domain-independent temporal stochastic simulator. We are using this first prototype to develop some applications, such as a robotic system controlled by classical planning and a logistics transportation system.
