PbP

Welcome to the homepage of PbP - a portfolio based planner.

Winner of the learning track of the 6th International Planning Competition (IPC6)

Winner of the learning track of the 7th International Planning Competition (IPC7)

The PbP Approach version 2.0 (June 2012)

PbP is an automated system that generates efficient domain-specific multi-planners from a portfolio of domain-independent planning techniques by:

  • (i) computing some sets of macro-actions for every planner in the portfolio;
  • (ii) optimizing the parameter setting of the parameterized planners in the portfolio;
  • (iii) selecting a promising combination of planners in the portfolio and relative useful macro-actions;
  • (iv) defining some running time slots for their round-robin scheduling during planning.

The configuration of the portfolio yielding the multi-planner relies on some knowledge about the performance of the planners and relative macro-actions, which is automatically generated from a training problem set.

We developed two versions of PbP:

  • PbP.s finds as quickly as possible a plan and then stops;
  • PbP.q finds a good quality plan using all the given CPU-time.

Download of PbP.s (executable code)

Download of PbP.q (executable code)

Main papers related to PbP

  1. Alfonso Gerevini, Alessandro Saetti, Mauro Vallati, An Automatically Configurable Portfolio-based Planner with Macro-actions: PbP, Proceedings of the 19th International Conference on Automated Planning & Scheduling (ICAPS-09), 19-23, AAAI, Thessaloniki (Greece), 2009.

  2. Alfonso Gerevini, Alessandro Saetti, Mauro Vallati, Learning and Exploiting Configuration Knowledge for a Portfolio-based Planner, Working notes of 19th International Conference on Automated Planning & Scheduling (ICAPS-09) - Workshop on Planning and Learning, Thessaloniki (Greece), 2009.

Planner Cluster table

IPC 6 domains - 2008

Table 1. Planner clusters, percentages of macros and macros definitions (in round brackets) selected by PbP for the IPC 6 domains.
More informations:

PTS stands for Planning Time Slots

Domain PbP.s PbP.q
  Planners & Macros PTS Conf. Planners & Macros PTS Conf.
Gold-miner YAHSP (CL) PTS Conf. LAMA-08 (BC), MacroFF (M1), LPG-td (0) PTS Conf.
Matching-BW ParLPG (--) params PTS Conf. Marvin (0), LAMA-08 (--), LPG-td (BC) PTS Conf.
N-puzzle ParLPG (--) params PTS Conf. Fast Downward (--), LAMA-08 (--), LPG-td (0) PTS Conf.
Parking MacroFF (M2) PTS Conf. Metric-FF (0), LAMA-08 (--) PTS Conf.
Sokoban ParLPG (--) params PTS Conf. MacroFF (M2), LPG-td (BC) PTS Conf.
Thoughtful Metric-FF (--), yahsp (--) PTS Conf. MacroFF (M5), Marvin (--), LAMA-08 (--) PTS Conf.

IPC 7 domains - 2011

Table 2. Planner clusters and percentages of macros (in round brackets) selected by PbP for the IPC 7 domains.
More informations:

PTS stands for Planning Time Slots

Domain PbP.s PbP.q
  Planners & Macros PTS Conf. Planners & Macros PTS Conf.
Barman SGPlan5 (BC) PTS Conf. Sgplan (CL), Metric-FF (--), LAMA-08 (--) PTS Conf.
Blocksworld ParLPG (BC) params PTS Conf. ParLPG (--) params, LPG-td (BC) PTS Conf.
Depots ParLPG (0) params, MacroFF (M2) PTS Conf. MacroFF (M2), ParLPG (0) params, Sgplan (CH) PTS Conf.
Gripper ParLPG (--) params PTS Conf. Marvin (--), ParLPG (--) params PTS Conf.
Parking MacroFF (M2) PTS Conf. Metric-FF (0), LAMA-08 (--) PTS Conf.
Rovers ParLPG (--) params PTS Conf. LAMA-08 (--), ParLPG (--) params PTS Conf.
Satellite ParLPG (--) params PTS Conf. SGPlan5 (), ParLPG (--) params, Marvin (0) PTS Conf.
Spanner ParLPG (--) params PTS Conf. LPG-td (--) PTS Conf.
TPP MacroFF (M1) PTS Conf. LAMA-08 (--), SGPlan5 (CH) PTS Conf.

All the planning systems excepting Wizard are used by PbP with the default parameters. The parameter configuration of Wizard are: bunching, chunking, and clumping.

PbP Team

Planning group coordinator:
Alfonso E. Gerevini
Current members:
Alfonso E. Gerevini, Alessandro Saetti, Mauro Vallati
Undergraduate students:
Beniamino Galvani (now graduated, previously involved).