A utomated planning is the process of finding an ordered sequence of actions that, starting from a given initial state, allows the transition to a state where a series of objectives are achieved. Actions are usually expressed in terms of preconditions and effects; that is, the requirements a state must meet for the action to be applied, and the changes subsequently made. Domain-independent planning relies on general problem-solving techniques to find an (approximately) optimal sequence of actions and has been the focus of numerous International Planning Competitions (IPCs) over the years.The first IPC was organized by Drew McDermott in 1998. For the following 10 years it was a biennial event and remains a keystone in the worldwide planning research community: the most recent, seventh, IPC took place in 2011. The major important contribution of the first competition was to establish a common standard language for defining planning problems -the planning domain definition language (PDDL) (McDermott 1998) -which has been developed and extended throughout the competition series. Today, the extended PDDL is still widely used and is key in allowing fair benchmarking of planners. Participation has increased dramatically over the years and a growing number of tracks have formed, representing the broadening community -see figure 1 for details. The three main tracks now operating are the deterministic, learning, and uncertainty tracks.The IPC has two main goals: to produce new benchmarks,

Recent discoveries in automated planning are broadening the scope of planners, from toy problems to real applications. However, applying automated planners to real-world problems is far from simple. On the one hand, the definition of accurate action models for planning is still a bottleneck. On the other hand, off-the-shelf planners fail to scale up and to provide good solutions in many domains. In these problematic domains, planners can exploit domain-specific control knowledge to improve their performance in terms of both speed and quality of the solutions. However, manual definition of control knowledge is quite difficult. This paper reviews recent techniques in machine learning for the automatic definition of planning knowledge. It has been organized according to the target of the learning process: automatic definition of planning action models and automatic definition of planning control knowledge. In addition, the paper reviews the advances in the related field of reinforcement learning.

No abstract

Current evaluation functions for heuristic planning are expensive to compute. In numerous planning problems these functions provide good guidance to the solution, so they are worth the expense. However, when evaluation functions are misguiding or when planning problems are large enough, lots of node evaluations must be computed, which severely limits the scalability of heuristic planners. In this paper, we present a novel solution for reducing node evaluations in heuristic planning based on machine learning. Particularly, we define the task of learning search control for heuristic planning as a relational classification task, and we use an off-the-shelf relational classification tool to address this learning task. Our relational classification task captures the preferred action to select in the different planning contexts of a specific planning domain. These planning contexts are defined by the set of helpful actions of the current state, the goals remaining to be achieved, and the static predicates of the planning task. This paper shows two methods for guiding the search of a heuristic planner with the learned classifiers. The first one consists of using the resulting classifier as an action policy. The second one consists of applying the classifier to generate lookahead states within a Best First Search algorithm. Experiments over a variety of domains reveal that our heuristic planner using the learned classifiers solves larger problems than state-of-the-art planners.

This paper describes our participation in the SemEval-2014 tasks 1, 3 and 10. We used an uniform approach for addressing all the tasks using the soft cardinality for extracting features from text pairs, and machine learning for predicting the gold standards. Our submitted systems ranked among the top systems in all the task and sub-tasks in which we participated. These results confirm the results obtained in previous SemEval campaigns suggesting that the soft cardinality is a simple and useful tool for addressing a wide range of natural language processing problems.