This paper provides theoretical and empirical comparisons of three recent hierarchical plan repair algorithms: SHOPFIXER, IPYHOPPER, and REWRITE. Our theoretical results show that the three algorithms correspond to three different definitions of the plan repair problem, leading to differences in the algorithms’ search spaces, the repair problems they can solve, and the kinds of repairs they can make. Understanding these distinctions is important when choosing a repair method for any given application. Building on the theoretical results, we evaluate the algorithms empirically in a series of benchmark planning problems. Our empirical results provide more detailed insight into the run- time repair performance of these systems and the coverage of the repair problems solved, based on algorithmic properties such as replanning, chronological backtracking, and back- jumping over plan trees.

We describe HTN‐MAKER, an algorithm for learning hierarchical planning knowledge in the form of task‐reduction methods for hierarchical task networks (HTNs). HTN‐MAKER takes as input a set of planning states from a classical planning domain and plans that are applicable to those states, as well as a set of semantically annotated tasks to be accomplished. The algorithm analyzes this semantic information to determine which portion of the input plans accomplishes a particular task and constructs task‐reduction methods based on those analyses. We present theoretical results showing that HTN‐MAKER is sound and complete. Our experiments in five well‐known planning domains confirm the theoretical results and demonstrate convergence toward a set of HTN methods that can be used to solve any problem expressible as a classical planning problem in that domain, relative to a set of goal types for which tasks have been defined. In three of the five domains, HTN planning with the learned methods scales much better than a modern classical planner.

In this paper we present a plan-plan distance metric based on Kolmogorov(Algorithmic) complexity. Generating diverse sets of plans is useful for task ssuch as probing user preferences and reasoning about vulnerability to cyberattacks. Generating diverse plans, and comparing different diverse planning approaches requires a domain-independent, theoretically motivated definition of the diversity distance between plans. Previously proposed diversity measures are not theoretically motivated, and can provide inconsistent results on the sameplans. We define the diversity of plans in terms of how surprising one plan is givenanother or, its inverse, the conditional information in one plan givenanother. Kolmogorov complexity provides a domain independent theory of conditional information. While Kolmogorov complexity is not computable, a related metric, Normalized Compression Distance (NCD), provides a well-behaved approximation. In this paper we introduce NCD as an alternative diversity metric, and analyze its performance empirically, in comparison with previous diversity measures, showing strengths and weaknesses of each.We also examine the use of different compressor sin NCD. We show how NCD can be used to select a training set for HTN learning,giving an example of the utility of diversity metrics. We conclude withsuggestions for future work on improving, extending, and applying it to serve new applications.

This paper describes a novel combination of Java program analysis and automated learning and planning architecture to the domain of Java vulnerability analysis. The key feature of our “HACKAR: Helpful Advice for Code Knowledge and Attack Resilience” system is its ability to analyze Java programs at development-time, identifying vulnerabilities and ways to avoid them. HACKAR uses an improved version of NASA’s Java PathFinder (JPF) to execute Java programs and identify vulnerabilities. The system features new Hierarchical Task Network (HTN) learning algorithms that (1) advance stateof-theart HTN learners with reasoning about numeric constraints, failures, and more general cases of recursion, and (2) contribute to problem-solving by learning a hierarchical dataflow representation of the program from the inputs of the program. Empirical evaluation demonstrates that HACKAR was able to suggest fixes for all of our test program suites. It also shows that HACKAR can analyze programs with string inputs that original JPF implementation cannot.

For HTN planning, we formally characterize and classify four kinds of problem spaces in which each node represents a planning problem or subproblem. Two of the problem spaces are searched by current HTN planning algorithms; the other two problem spaces are new.This enables us to provide:Sufficient (and in one case, necessary) conditions for finiteness of each kind of problem space. The conditions can be evaluated up-front to see if an HTN planning problem is finite.Loop-detection tests that can be used in HTN planners to ensure termination when the problem space is finite.A way to compute the correct value for an upper-bound parameter in an HTN-to-PDDL translation algorithm published in IJCAI-2009.Planning algorithms that utilize the two new problem spaces to guarantee termination on broader classes of planning problems than previous HTN planning algorithms.

We present a Bayesian approach to learning flexible safety constraints and subsequently verifying whether plans satisfy these constraints. Our approach, called the Safety Constraint Learner/Checker (SCLC), infers safety constraints from a single expert demonstration trace and minimal background knowledge, and applies these constraints to the solutions proposed by multiple planning agents in an integrated and heterogeneous ensemble. The SCLC calculates how much to blame plan fragments (partial solutions) generated by the individual planning agents. This information is used when composing these fragments into a final overall plan. In particular, fragments whose safety violations exceed a threshold are rejected. This facilitates the generation of safe plans. We have integrated the SCLC within the Generalized Integrated Learning Architecture, which was designed for Defense Advanced Research Projects Agency (DARPA)’s Integrated Learning (IL) program. The main goal of the IL program is to promote the development and success of sophisticated systems that learn to solve challenging real‐world problems based on a simple demonstration by a human expert and exiguous domain knowledge. We present experimental results showing the advantages of the SCLC on two multiagent problem‐solving tasks that were benchmark applications in DARPA’s IL program.

We present a novel ensemble architecture for learning problem-solving techniques from a very small number of expert solutions and demonstrate its effectiveness in a complex real-world domain. The key feature of our “Generalized Integrated Learning Architecture” (GILA) is a set of heterogeneous independent learning and reasoning (ILR) components, coordinated by a central meta-reasoning executive (MRE). The ILRs are weakly coupled in the sense that all coordination during learning and performance happens through the MRE. Each ILR learns independently from a small number of expert demonstrations of a complex task. During performance, each ILR proposes partial solutions to subproblems posed by the MRE, which are then selected from and pieced together by the MRE to produce a complete solution. The heterogeneity of the learner-reasoners allows both learning and problem solving to be more effective because their abilities and biases are complementary and synergistic. We describe the application of this novel learning and problem solving architecture to the domain of airspace management, where multiple requests for the use of airspaces need to be deconflicted, reconciled, and managed automatically. Formal evaluations show that our system performs as well as or better than humans after learning from the same training data. Furthermore, GILA outperforms any individual ILR run in isolation, thus demonstrating the power of the ensemble architecture for learning and problem solving.

We consider how to learn Hierarchical Task Networks (HTNs) for planning problems in which both the quality of solution plans generated by the HTNs and the speed at which those plans are found is important. We describe an integration of HTN Learning with Reinforcement Learning to both learn methods by analyzing semantic annotations on tasks and to produce estimates of the expected values of the learned methods by performing Monte Carlo updates. We performed an experiment in which plan quality was inversely related to plan length. In two planning domains, we evaluated the planning performance of the learned methods in comparison to two state-of-the-art satisficing classical planners, FastForward and SGPlan6, and one optimal planner, HSP*. The results demonstrate that a greedy HTN planner using the learned methods was able to generate higher quality solutions than SGPlan6 in both domains and FastForward in one. Our planner, FastForward, and SGPlan6 ran in similar time, while HSP* was exponentially slower.

In this article, we describe a new approach that gives an explicit probabilistic interpretation for social networks. In particular, we focus on the observation that many existing Web-based trust-inference algorithms conflate the notions of “trust” and “confidence,” and treat the amalgamation of the two concepts to compute the trust value associated with a social relationship. Unfortunately, the result of such an algorithm that merges trust and confidence is not a trust value, but rather a new variable in the inference process. Thus, it is hard to evaluate the outputs of such an algorithm in the context of trust inference. This article first describes a formal probabilistic network model for social networks that allows us to address that issue. Then we describe SUNNY, a new trust inference algorithm that uses probabilistic sampling to separately estimate trust information and our confidence in the trust estimate and use the two values in order to compute an estimate of trust based on only those information sources with the highest confidence estimates. We present an experimental evaluation of SUNNY. In our experiments, SUNNY produced more accurate trust estimates than the well-known trust inference algorithm TidalTrust, demonstrating its effectiveness. Finally, we discuss the implications these results will have on systems designed for personalizing content and making recommendations.

This paper describes a system for learning domain-dependent knowledge in the form of goal-indexed Hierarchical Task Networks (HTNs). DLIGHT is a goal-directed incremental learning algorithm which observes solution traces and generates rules for solving problems. One of the main challenges in learning this kind of knowledge is determining a good level of generality. Analytical methods, such as explanation-based macro-operator learning, construct very specific structures that guarantee a successful execution when applicable but generalize poorly to new problems. Previous goal-directed learning approaches produce hierarchical rules with more relaxed preconditions, but the learned knowledge suffers from over-generality. Our approach builds on one such approach but it strikes a better balance between generality and specificity. This is done by carrying out a goal-dependency analysis to determine the structure of the hierarchy and precondition of each rule to follow the successful solutions more closely while maintaining generality. We hypothesize that this algorithm produces HTNs that generalize well and can solve problems efficiently. We evaluate the system's behavior experimentally in several planning scenarios and conclude with related work and future research paths.

Despite the recent advances in planning for classical domains, the question of how to use domain knowledge in planning is yet to be completely and clearly answered. Some of the existing planners use domain-independent search heuristics, and some others depend on intensively-engineered domain-specific knowledge to guide the planning process. In this paper, we describe an approach to combine ideas from both of the above schools of thought. We present Duet, our planning system that incorporates the ability of using hierarchical domain knowledge in the form of Hierarchical Task Networks (HTNs) as in SHOP2 [14] and using domain-independent local search techniques as in LPG [8]. In our experiments, Duet was able to solve much larger problems than LPG could solve, with only minimal domain knowledge encoded in HTNs (much less domain knowledge than SHOP2 needed to solve those problems by itself).

The SHOP2 planning system received one of the awards for distinguished performance in the 2002 International Planning Competition. This paper describes the features of SHOP2 which enabled it to excel in the competition, especially those aspects of SHOP2 that deal with temporal and metric planning domains.