We propose an approach to remodelling classical planning domains via the addition of macro operators and removal of original operators either for the domain as a whole or instance-by-instance. For the latter remodelling, we train a predictor to choose the best reformulation of the domain based on instance characteristic. In the domain level remodelling, we try find a fixed remodelling that works best on average over our training set. Operator removal does not generally preserve solubility and proving solubility preservation of domain models is PSPACE-complete. So we use an approach that uses training instances to empirically estimate the probability of solubility preservation and maintains a minimum value of that probability on the training instances. We show that the instance-specific approach outperforms the traditional best-on-average macro-only remodelling approach in 9 out of 14 cases of domain/macro-source combinations, and that it can outperform fixed domain-based models generated with existing macro learning tools.

Unlike mathematical programming and SAT solving, ConstraintProgramming (CP) is based on the idea that both modeling and solving of combinatorial optimization problems can be based on conjunctions of loosely coupled, recurring, combinatorial sub-problems (aka "global constraints"). This rich representational approach means that, for better or for worse, pretty much anything can be expressed as a global constraint. Much of CP's success, however, has come from exploiting only one aspect of the rich constraint definition: global constraint propagation. In this talk, I will investigate how work in CP, SAT, AI planning, and mathematical programming can be understood as more seriously pursuing the implications of a rich constraint definition and how the interplay between local and global information can lead to dynamic problem reformulations and a flexible hybrid solver architecture.

This paper addresses the Target-Value Search (TVS) problem, which is the problem of finding a path between two nodes in a graph whose cost is as close as possible to a given target value T. This problem has been previously addressed only for directed acyclic graphs. In this work we develop the theory required to solve this problem optimally for any type of graphs. We modify traditional heuristic search algorithms for this setting, and propose a novel bidirectional search algorithm that is specifically suited for TVS. The benefits of this bidirectional search algorithm are discussed both theoretically and experimentally on several domains. A longer version of this work was accepted to IJCAI-2013 (Linares Lopez et al. 2013)

We consider multi-agent planning in which the agents' optimal plans are solutions to mixed-integer programs (MIP) that are coupled via integer constraints. While in principle, one could find the joint solution by combining the separate problems into one large joint centralized MIP, this approach rapidly becomes intractable for growing numbers of agents and large problem domains. To address this issue, we propose an iterative approach that combines conflict detection with constraint-generation whereby the agents plan repeatedly until all conflicts are resolved. In each planning iteration, the agents plan with as few other agents and interaction-constraints as possible. This yields an optimal method that can reduce computation markedly. We test our approach in the context of multi-agent collision avoidance in graphs with indivisible flows. Our initial simulations on randomized graph routing problems confirm predicted optimality and reduced computational effort.

The Increasing Cost Tree Search (ICTS) algorithm is used to produce optimal solutions to the multi-agent path finding problem (MAPF). In this problem, multiple agents are trying to reach their goals without conflicting with each other, while minimizing the total cost of the paths. ICTS has been shown to be very effective in finding optimal solutions. In this paper we consider the problem of finding solutions with bounded suboptimality by changing the order in which ICTS searches its increasing cost tree. With a variety of strategies, we are unable to consistently and significantly reduce the cost of ICTS. Further experimentation suggests why significantly more work is needed to modify ICTS to find suboptimal solutions.

In this paper, we consider the problem of planning any-angle paths with small numbers of turns on grid maps. We propose a novel heuristic search algorithm called Link* that returns paths containing fewer turns at the cost of slightly longer path lengths. Experimental results demonstrate that Link* can produce paths with fewer turns than other any-angle path planning algorithms while still maintaining comparable path lengths. Because it produces this type of path, artificial agents can take advantage of Link* when the cost of turns is expensive.

In this paper, we present a conceptually simple, easy-to-implement real-time search algorithm suitable for a priori partially known environments. Instead of performing a series of searches towards the goal, like most Real-Time Heuristic Search Algorithms do, our algorithm follows the arcs of a tree T rooted in the goal state that is built initially using the heuristic h. When the agent observes that an arc in the tree cannot be traversed in the actual environment, it removes such an arc from T and our algorithm carries out a reconnection search whose objective is to find a path between the current state and any node in T. The reconnection search need not be guided by $h$, since the search objective is not to encounter the goal. Furthermore, h need not be updated. We implemented versions of our algorithm that utilize various blind search algorithms for reconnection. We show experimentally that these implementations significantly outperform state-of-the-art real-time heuristic search algorithms for the task of pathfinding in grids. In grids, our algorithms, which do not incorporate any geometrical knowledge, naturally behaves similarly to a bug algorithm, moving around obstacles, and never returning to areas that have been visited in the past. In addition, we prove theoretical properties of the algorithm.

Multi-agent pathfinding is an important challenge that relates to combinatorial search and has many applications, such as warehouse management, robotics and computer games. Finding an optimal solution is NP-hard and raises scalability issues for optimal solvers. Interestingly, however, it takes linear time to check the feasibility of an instance. These linear-time feasibility tests can be extended to provide path planners but to the best of the authors’ knowledge no such solver has been provided for general graphs. This work first describes a path planner that is inspired by a linear-time feasibility test for multi-agent pathfinding on general graphs. Initial experiments indicated reasonable scalability but worse path quality relative to existing suboptimal solutions. This led to the development of an algorithm that achieves both efficient running time and path quality relative to the alternatives and which finds a solution on available benchmarks. The paper outlines the relation of the final method to the feasibility tests and existing suboptimal planners. Experimental results evaluate the different algorithms, including an optimal solver.

In this paper we advance the line of research launched by Knuth which was later improved by Chen for predicting the size of the search tree expanded by heuristic search algorithms such as IDA*. Chen's Stratified Sampling (SS) uses a partition of the nodes in the search tree called type system to guide its sampling. Recent work has shown that SS using type systems based on integer-valued heuristic functions can be quite effective. However, type systems based on real-valued heuristic functions are often too large to be practical. We use the k-means clustering algorithm for creating effective type systems for domains with real-valued heuristics. Orthogonal to the type systems, another contribution of this paper is the introduction of an algorithm called Active SS. SS allocates the same number of samples for each type. Active SS is the application of the idea of active sampling to search trees. Active SS allocates more samples to the types with higher uncertainty. Our empirical results show that (i) SS using clustering-based type systems tends to produce better predictions than competing schemes that do not use a type system, and that (ii) Active SS can produce better predictions than the regular version of SS.

In multiobjective state space graph problems, each solution-path is evaluated by a cost vector. These cost vectors can be partially or completely ordered using a preference relation compatible with Pareto dominance. In this context, multiobjective preference-based search (MOPBS) aims at computing the preferred feasible solutions according to a predefined preference model, these preferred solutions being a subset (possibly the entire set) of Pareto optima. Standard algorithms for MOPBS perform a unidirectional search developing the search tree forward from the initial state to a goal state. Instead, in this paper, we focus on bidirectional search algorithms developing simultaneously one forward and one backward search tree. Although bi-directional search has been tested in various single objective problems, its efficiency in a multiobjective setting has never been studied. In this paper, we present several implementations of bidirectional preference-based search convenient for the multiobjective case and investigate their efficiency.