Abstract In this article, we present an overview of the 2023 International Planning Competition. It featured five distinct tracks designed to assess cutting‐edge methods and explore frontiers planning within these settings: classical (deterministic) track, numeric Hierarchical Task Networks (HTN) learning probabilistic reinforcement track. Each evaluated methodologies through one or more subtracks, with goal pushing boundaries current planner performance. To achieve objective, competition...

In this paper, we focus on the inference of mutex groups in lifted (PDDL) representation. We formalize and prove that most commonly used translator from Fast Downward (FD) planning system infers a certain subclass groups, called fact-alternating (fam-groups). Based that, show previously proposed fam-groups-based pruning techniques for STRIPS representation can be utilized during grounding process with fam-groups, i.e., before full is known. Furthermore, propose an improved algorithm...

Polynomial-time heuristic functions for planning are commonplace since 20 years. But polynomial-time in which input? Almost all existing approaches based on a grounded task representation, not the actual PDDL input is exponentially smaller. This limits practical applicability to cases where representation "small enough". Previous attempts tackle this problem delete relaxation leveraged symmetries reduce blow-up. Here we take more radical approach, applying an additional obtain function that...

Checking whether action effects can be undone is an important question for determining, instance, a planning task has dead-ends. In this paper, we investigate the reversibility of actions, that is, when reverted by applying other in order to return original state. We propose broad notion generalizes previously defined versions and interesting properties relevant restrictions. particular, concept uniform guarantees independently state which was applied, using so-called reverse plan. addition,...

Mutex groups are defined in the context of STRIPS planning as sets facts out which, maximally, one can be true any state reachable from initial state. The importance computing and exploiting mutex was repeatedly pointed many studies. However, theoretical analysis is sparse current literature. This work provides a complexity showing that inference hard itself (PSPACE-Complete) it also shows tight relationship between graph cliques. result motivates us to propose new type group called...

Heuristics are a crucial component in modern planning systems. In optimal multiagent the state of art is to compute heuristic locally using only information available single agent. This approach has major deficiency as local shortest path can arbitrarily underestimate true cost global problem. As solution, we propose distributed version state-of-the-art LM-Cut heuristic. We show that our algorithm provides estimates provably equal centralized computed on also evaluate experimentally and...

Simplifying classical planning tasks by removing operators while preserving at least one optimal solution can significantly enhance the performance of planners. In this paper, we introduce notion operator mutex, which is a set that cannot all be part same (strongly) plan. We propose four different methods for inference mutexes and experimentally verify they found in sizable number tasks. show how used combination with structural symmetries to safely remove from task.

Testing is a promising way to gain trust in neural action policies π. Previous work on policy testing sequential decision making targeted environment behavior leading failure conditions. But if the unavoidable given that behavior, then π not actually blame. For situation qualify as "bug" π, there must be an alternative π' does better. We introduce generic framework based intuition. This raises bug confirmation problem, deciding whether or state bug. analyze use of optimistic and pessimistic...

Potential heuristics assign a numerical value (potential) to each fact and compute the heuristic for given state as sum of these potentials. A mutex is an invariant stating that certain combination facts cannot be part any reachable state. In this paper, we use mutexes improve potential in two ways. First, show mutex-based disambiguations goal preconditions operators leads less constrained linear program yielding stronger heuristics. Second, utilize construction new optimization functions...

Distributed heuristic search is a well established technique for multi-agent planning. It has been shown that distributed heuristics may crucially improve the guidance, but are costly in terms of communication and computation time. One solution to compute additively, sense each agent can its part independently obtain complete estimate by summing up individual parts. In this paper, we show recently published potential good candidate such heuristic, moreover admissible. We also demonstrate how...

A lifted mutex group is a schematic first-order description of sets facts such that each set contains out which at most one can hold in any reachable state. It was previously shown groups be used for pruning operators during grounding PDDL tasks, i.e., it possible to prune unreachable and dead-end even before the grounded representation known. Here, we show applying technique does not require modification procedure. Instead, compile conditions under use directly into preconditions actions on...

Classical planning tasks are usually modelled in the PDDL which is a schematic language based on first-order logic. Nevertheless, most of current planners turn this representation into propositional one via grounding process. It well known that process may cause an exponential blowup. Therefore it important to detect grounded atoms redundant sense they not necessary for finding plan and therefore does need generate them. This done by relaxed reachability analysis, can be improved employing...

Classical planning tasks are modelled in PDDL which is a schematic language based on first-order logic. Most of the current planners turn this lifted representation into propositional one via grounding process. However, may cause an exponential blowup. Therefore it important to investigate methods for searching plans level. To build state-based planner, necessary invent heuristics. We introduce maps between preserving allowing transform task smaller one. propose novel method computing...

Symbolic search, using Binary Decision Diagrams (BDDs) to represent sets of states, is a competitive approach optimal planning. Yet heuristic search in this context remains challenging. The many advances on admissible planning heuristics are not directly applicable, as they evaluate one state at time. Indeed, progress functions symbolic has been limited and even very informed have shown be detrimental. Here we show how connection can made stronger for LP-based potential heuristics. Our key...

Privacy-Preserving Multi-Agent Planning (PP-MAP) has recently gained the attention of research community, resulting in a number PP-MAP planners and theoretical works. Many such lack strong guarantees, thus order to compare their abilities w.r.t. privacy, versatile practical metric is crucial. In this work, we propose metric, building on existing work. We generalize implement approach be applicable real planning domains provide an evaluation stateof-the-art over standard set benchmarks. The...

Testing was recently proposed as a method to gain trust in learned action policies classical planning. Test cases this setting are states generated by fuzzing process that performs random walks from the initial state. A bias attempts these towards policy bugs, is, where sub-optimally. Prior work explored simple based on policy-trace cost. Here, we investigate topic more deeply. We introduce three new biases analyses of shape, estimating whether trace is close looping back itself, it contains...

Potential heuristics assign numerical values (potentials) to state features, where each feature is a conjunction of facts. It was previously shown that the informativeness potential can be significantly improved by considering complex but computing potentials over all pairs facts already too costly in practice. In this paper, we investigate whether using just few high-dimensional features instead conjunctions up dimension n result while keeping computational cost at bay. We focus on (a)...

Heuristic search guides the exploration of states via heuristic functions h estimating remaining cost. Symbolic instead replaces individual with that state sets, compactly represented using binary decision diagrams (BDDs). In cost-optimal planning, explicit performs best overall, but symbolic in many domains, so both approaches together constitute art. Yet combinations two have far not been an unqualified success, because (i) must be applicable to sets rather than ones, and (ii) different...