The IBaCoP Planning System: Instance-Based Configured Portfolios

Abstract

Sequential planning portfolios are very powerful in exploiting the complementary strength of different automated planners. The main challenge of a portfolio planner is to define which base planners to run, to assign the running time for each planner and to decide in what order they should be carried out to optimize a planning metric. Portfolio configurations are usually derived empirically from training benchmarks and remain fixed for an evaluation phase. In this work, we create a per-instance configurable portfolio, which is able to adapt itself to every planning task. The proposed system pre-selects a group of candidate planners using a Pareto-dominance filtering approach and then it decides which planners to include and the time assigned according to predictive models. These models estimate whether a base planner will be able to solve the given problem and, if so, how long it will take. We define different portfolio strategies to combine the knowledge generated by the models. The experimental evaluation shows that the resulting portfolios provide an improvement when compared with non-informed strategies. One of the proposed portfolios was the winner of the Sequential Satisficing Track of the International Planning Competition held in 2014.

Deep Learning for Cost-Optimal Planning: Task-Dependent Planner Selection

“…While some portfolios settle on a schedule offline (Helmert et al 2011;Núñez, Borrajo, and Linares López 2014;Seipp, Sievers, and Hutter 2014a;2014b;2014c), i.e., ahead of execution, others try to select good planners online based on the given task (Cenamor, de la Rosa, and Fernández 2016;. The latter is based on machine learning techniques, training a classifier on planning instances, represented as a vector of hand-crafted features.…”

Determining Representativeness of Training Plans: A Case of Macro-Operators

“…Complementary to developing planning engines, the learning for planning research area aims at learning useful knowledge about a class of planning tasks that can be exploited for improving the performance of planning engines. Examples of learning for planning include the configuration of portfolios, such as PbP [2] or IBACOP [3], that delivered outstanding performance in recent IPCs.…”

Static and Dynamic Portfolio Methods for Optimal Planning: An Empirical Analysis

“…Among the latter, we can identify two main classes: static portfolios, which run the same schedule of planners on every given problem instance, and portfolios based on per-instance planner schedules. Cedalion [9] and Fast Downward Stone Soup [10] are well-known examples of static portfolio-based planners, while IBaCoP [11] selects the best planner schedule on a per-instance basis [12]. Here, we introduce a third class, that of dynamic portfolios, comprised of planners in which the schedule is created dynamically, during execution, based on performance data from earlier runs of the given planners as well as on features of the planning instance to be solved.…”