Hierarchical Task Network Research Articles

Change the Plan — How Hard Can That Be?

Interaction with users is a key capability of planning systems that are applied in real-world settings. Such a system has to be able to react appropriately to requests issued by its users. Most of these systems are based on a generated plan that is continually criticised by him, resulting in a mixed-initiative planning system. We present several practically relevant requests to change a plan in the setting of hierarchical task network planning and investigate their computational complexity. On the one hand, these results provide guidelines when constructing algorithms to execute the respective requests, but also provide translations to other well-known planning queries like plan existence or verification. These can be employed to extend an existing planner such that it can form the foundation of a mixed-initiative planning system simply by adding a translation layer on top.

Unsupervised Learning of HTNs in Complex Adversarial Domains

While Hierarchical Task Networks are frequently cited as flexible and powerful planning models, they are often ignored due to the intensive labor cost for experts/programmers, due to the need to create and refine the model by hand. While recent work has begun to address this issue by working towards learning aspects of an HTN model from demonstration, or even the whole framework, the focus so far has been on simple domains, which lack many of the challenges faced in the real world such as imperfect information and real-time environments. I plan to extend this work using the domain of real-time strategy (RTS) games, which have gained recent popularity as a challenging and complex domain for AI research.

Dynamic Aircraft Carrier Flight Deck Task Planning based on HTN

In the operating environment of naval aircraft carrier, the move towards an increasing unmanned operational tempo with a reduced manning environment will request more automation in the planning and scheduling of aircraft, ground vehicles, and crew. This paper investigates the aircraft carrier deck operations planning problem based on Hierarchical Task Network (HTN) planning method. An HTN based dynamic operations planning process is proposed. A feasible action plan could be rapidly derived to complete a set of takeoff and landing tasks with the constraints of flight deck layout and limited resources such as catapults, elevators, and operators, etc. Corresponding to the newly generated tasks, the replanning process is triggered to cope with the changing conditions in highly dynamic environments.

Hierarchical Task Planning for Multiarm Robot with Multiconstraint

Multiarm systems become the trends of space robots, for the on-orbit servicing missions are becoming more complex and various. A hierarchical task planning method with multiconstraint for multiarm space robot is presented in this paper. The process of task planning is separated into two hierarchies: mission profile analysis and task node planning. In mission profile analysis, several kinds of primitive tasks and operators are defined. Then, a complex task can be decomposed into a sequence of primitive tasks by using hierarchical task network (HTN) with those primitive tasks and operators. In task node planning,A⁎algorithm is improved to adapt the continuous motion of manipulator. Then, some of the primitive tasks which cannot be executed directly because of constraints are further decomposed into several task nodes by using improvedA⁎algorithm. Finally, manipulators execute the task by moving from one node to another with a simple path plan algorithm. The feasibility and effectiveness of the proposed task planning method are verified by simulation.

Graph-based motor primitive generation framework

AbstractUnmanned aerial vehicles (UAVs) have many potential applications, such as delivery, leisure, and surveillance. To enable these applications, making the UAVs fly autonomously is the key issue, and requires defining UAV motor primitives. Diverse attempts have been made to automatically generate motor primitives for UAVs and robots. However, given that UAVs usually do not fly as expected because of external environmental factors, a novel approach for UAVs needs to be designed. This paper proposes a demonstration-based method that generates a graph-based motor primitive. In the experiment, an AR.Drone 2.0 was utilized. By controlling the AR.Drone 2.0, four motor primitives are generated and combined as a graph-based motor primitive. The generated motor primitives can be performed by a planner or a learner, such as a hierarchical task network or Q-learning. By defining the executable conditions of the motor primitives based on measured properties, the movements of the graph-based motor primitive can be chosen depending on changes in the indoor environment.

Decision-making model to generate novel emergency response plans for improving coordination during large-scale emergencies

Developing joint emergency response plans is an effective method to coordinate multi-agency response endeavors. This study presents a novel emergency response plan structure that considers emergency command operation requirements, such as explicitly expressing the incident objective decomposition structure, formalizing decisions in a context-sensitive manner, supporting the synchronization of responding activities with variable interval, encoding complex temporal constraints, and providing temporal flexibility. A decision-making model is developed to generate these domain-specific action plans automatically based on integrating hierarchical task network (HTN) planning and scheduling technologies. This model presents several valuable contributions to existing state-based forwarding HTN planning paradigms. First, an enhanced HTN is designed to record traversed HTN exploration space for constructing of incident objective decomposition structure and decision-making contexts. Second, the model generates temporal flexible action plans that enable the handling of temporal uncertainty in the emergency response domain. A novel concurrency controlling mechanism to ensure the parallelism of response activities with variable intervals is also proposed based on the temporally enhanced planning state that represents a dynamic emergency situation. Finally, the proposed model explicitly represents the starting and ending time of all tasks in the task network to provide complete temporal flexibility. In particular, a dedicated temporal management method taking full advantage of the decomposition structure induced by the HTN planning process is proposed for propagating time constraints on the underling Simple Temporal Network (STN) incrementally. An empirical study on typhoon evacuation demonstrates that the presented model is suitable for solving real-world problems. Therefore, the decision-making model can be applied as a computational model for the development process of emergency response plans, and will be embedded as a reasoning logic in an emergency command decision support system.

Search Complexities for HTN Planning

Abstract : Hierarchical Task Network (HTN) planning is the problem of decomposing an initial task into a sequence of executable steps. Often viewed as just a way to encode human knowledge to solve classical planning problems faster, HTN planning is more expressive than classical planning, even to the point of being undecidable in the general case. However, HTN planning is not just a way to solve planning problems faster, but is itself a search problem that can benefit from its own distinct search algorithms and heuristics. The dissertation examines the complexities of various HTN planning problem classes in order to motivate the development of heuristic search algorithms for HTN planning which are guaranteed to terminate on a large class of syntactically identifiable problems, as well as domain independent heuristics for those algorithms to use. This will allow HTN planning to be used in a number of areas where problems may be unsolvable, including during the initial development of a domain and for use in policy generation in non-deterministic planning environments. In particular, this dissertation analyzes two commonly used algorithms for HTN planning and describes the subsets of HTN problems that these algorithms terminate on. This allows us to discuss the run-times of these algorithms and compare the expressivity of the classes of problems they decide. We provide two new HTN algorithms which terminate on a strictly broader and more expressive set of HTN problems. We also analyze the complexity of delete-free HTN planning, an analogue to delete-free classical planning which is the base of many classical planning heuristics. We show that delete-free HTN planning is NP-complete, putting the existence of strict-semantics delete-relaxation-based HTN heuristics out of reach for practical purposes.

Intent Understanding Using an Activation Spreading Architecture

In this paper, we propose a new approach for recognizing intentions of humans by observing their activities with a color plus depth (RGB-D) camera. Activities and goals are modeled as a distributed network of inter-connected nodes in an Activation Spreading Network (ASN). Inspired by a formalism in hierarchical task networks, the structure of the network captures the hierarchical relationship between high-level goals and low-level activities that realize these goals. Our approach can detect intentions before they are realized and it can work in real-time. We also extend the formalism of ASNs to incorporate contextual information into intent recognition. We further augment the ASN formalism with special nodes and synaptic connections to model ordering constraints between actions, in order to represent and handle partial-order plans in our ASN. A fully functioning system is developed for experimental evaluation. We implemented a robotic system that uses our intent recognition to naturally interact with the user. Our ASN based intent recognizer is tested against three different scenarios involving everyday activities performed by a subject, and our results show that the proposed approach is able to detect low-level activities and recognize high-level intentions effectively in real-time. Further analysis shows that contextual and partial-order ASNs are able to discriminate between otherwise ambiguous goals.

Tight Bounds for HTN Planning

Although HTN planning is in general undecidable, there are many syntactically identifiable sub-classes of HTN problems that can be decided. For these sub-classes, the decision procedures provide upper complexity bounds. Lower bounds were often not investigated in more detail, however. We generalize a propositional HTN formalization to one that is based upon a function-free first-order logic and provide tight upper and lower complexity results along three axes: whether variables are allowed in operator and method schemas, whether the initial task and methods must be totally ordered, and where recursion is allowed (arbitrary recursion, tail-recursion, and acyclic problems). Our findings have practical implications, both for the reuse of classical planning techniques for HTN planning, and for the design of efficient HTN algorithms.

On the Complexity of HTN Plan Verification and Its Implications for Plan Recognition

In classical planning it is easy to verify if a given sequence of actions is a solution to a planning problem. It has to be checked whether the actions are applicable in the given order and if a goal state is reached after executing them. In this paper we show that verifying whether a plan is a solution to an HTN planning problem is much harder. More specifically, we prove that this problem is NP-complete, even for very simple HTN planning problems. Furthermore, this problem remains NP-complete if an executable sequence of tasks is already provided. HTN-like hierarchical structures are commonly used to represent plan libraries in plan and goal recognition. By applying our result to plan and goal recognition we provide insight into its complexity.

Integrating task planning, execution and monitoring for a domestic service robot

Abstract Embodied artificial agents operating in dynamic, real-world environments need architectures that support the special requirements that exist for them. Architectures are not always designed from scratch and the system then implemented all at once, but rather, a step-wise integration of components is often made to increase functionality. In order to increase flexibility and robustness, a task planner was integrated into an existing architecture and the planning process was coupled with the pre-existing execution and the basic monitoring processes. This involved the conversion of monolithic SMACH scenario scripts (state-machine execution scripts) into modular states that can be called dynamically based on the plan that was generated by the planning process. The procedural knowledge encoded in such state machines was used to model the planning domain for two RoboCup@Home scenarios on a Care-O-Bot 3 robot. This was done for the JSHOP2 hierarchical task network (HTN) planner. A component which iterates through a generated plan and calls the appropriate SMACH states was implemented, thus enabling the scenarios. Crucially, individual monitoring actions which enable the robot to monitor the execution of the actions were designed and included, thus providing additional robustness.

모듈형 행동선택네트워크를 이용한 거울뉴런과 마음이론 기반의 의도대응 모델

Although service robots in various fields are being commercialized, most of them have problems that depend on explicit commands by users and have difficulty to generate robust reactions of the robot in the unstable condition using insufficient sensor data. To solve these problems, we modeled mirror neuron and theory of mind systems, and applied them to a robot agent to show the usefulness. In order to implement quick and intuitive response of the mirror neuron, the proposed intention-response model utilized behavior selection networks considering external stimuli and a goal, and in order to perform reactions based on the long-term action plan of theory of mind system, we planned behaviors of the sub-goal unit using a hierarchical task network planning, and controled behavior selection network modules. Experiments with various scenarios revealed that appropriate reactions were generated according to external stimuli.

Process Planner: An Approach to Conceptualize Business Practices in Private Cloud

The expansion of business is majorly driven by client and quality oriented architecture. Every commercial setup has its own business pattern and process of implementation. Large numbers of alternatives are supporting trade in terms of business formation, business development, and business expansion also, sometimes in each of its incremental stage. For business development, flexible technology is required, that can be a part of an existing technology and that would help to model business based on business rules. Business concepts need to be well organized and predefined for collaboration to ensure complete client satisfaction. This paper proposes an architecture that is light weight in terms of resource requirements that is modeled on the concept-building approach which can further embedded into private or public cloud to support service oriented architecture. Inputs related to application are taken from a collaborating client, which is then forwarded to the planner for matching. The discovered web services (belonging to other clients) are then chosen for further collaboration. The planner relies on a concept-based approach, rather than using complex database schema. Concept of Artificial Intelligence (AI), a Hierarchical Task Network (HTN) play an important role in forming and analyzing the concept built. Each of the tasks is decomposed till depth (primary node), on the base of filter parameters mentioned by client. Medi-Care scenario is considered to show applicability of the approach. Approach solves and formulates the problem while using client criteria of search to map required output.

HACKAR: Helpful Advice for Code Knowledge and Attack Resilience

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.

Leveraging Multiple Artificial Intelligence Techniques to Improve the Responsiveness in Operations Planning: ASPEN for Orbital Express

The challenging timeline for DARPA's Orbital Express mission demanded a flexible, responsive, and (above all) safe approach to mission planning. Mission planning for space is challenging because of the mixture of goals and constraints. Every space mission tries to squeeze all of the capacity possible out of the spacecraft. For Orbital Express, this means performing as many experiments as possible, while still keeping the spacecraft safe. Keeping the spacecraft safe can be very challenging because we need to maintain the correct thermal environment (or batteries might freeze), we need to avoid pointing cameras and sensitive sensors at the sun, we need to keep the spacecraft batteries charged, and we need to keep the two spacecraft from colliding … made more difficult as only one of the spacecraft had thrusters.Because the mission was a technology demonstration, pertinent planning information was learned during actual mission execution. For example, we didn't know for certain how long it would take to transfer propellant from one spacecraft to the other, although this was a primary mission goal. The only way to find out was to perform the task and monitor how long it actually took. This information led to amendments to procedures, which led to changes in the mission plan. In general, we used the ASPEN planner scheduler to generate and validate the mission plans. ASPEN is a planning system that allows us to enter all of the spacecraft constraints, the resources, the communications windows, and our objectives. ASPEN then could automatically plan our day. We enhanced ASPEN to enable it to reason about uncertainty. We also developed a model generator that would read the text of a procedure and translate it into an ASPEN model. Note that a model is the input to ASPEN that describes constraints, resources, and activities. These technologies had a significant impact on the success of the Orbital Express mission. Finally, we formulated a technique for converting procedural information to declarative information by transforming procedures into models of hierarchical task networks (HTNs). The impact of this effort on the mission was a significant reduction in (1) the execution time of the mission, (2) the daily staff required to produce plans, and (3) planning errors. Not a single misconfigured command was sent during operations.

Learning Hierarchical Task Models from Input Traces

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.

Finding Ways to Get the Job Done: An Affordance-Based Approach

Adapting plans to changes in the environment by finding alternatives and taking advantage of opportunities is a common human behavior. The need for such behavior is often rooted in the uncertainty produced by our incomplete knowledge of the environment. While several existing planning approaches deal with such issues, artificial agents still lack the robustness that humans display in accomplishing their tasks. In this work, we address this brittleness by combining Hierarchical Task Network planning, Description Logics, and the notions of affordances and conceptual similarity. The approach allows a domestic service robot to find ways to get a job done by making substitutions. We show how knowledge is modeled, how the reasoning process is used to create a constrained planning problem, and how the system handles cases where plan generation fails due to missing/unavailable objects. The results of the evaluation for two tasks in a domestic service domain show the viability of the approach in finding and making the appropriate goal transformations.

On the Feasibility of Planning Graph Style Heuristics for HTN Planning

In classical planning, the polynomial-time computability of propositional delete-free planning (planning with only positive effects and preconditions) led to the highly successful Relaxed Graphplan heuristic. We present a hierarchy of new computational complexity results for different classes of propositional delete-free HTN planning, with two main results: We prove that finding a plan for the delete-relaxation of a propositional HTN problem is NP-complete: hence unless P=NP, there is no directly analogous GraphPlan heuristic for HTN planning. However, a further relaxation of HTN planning (delete-free HTN planning with task insertion) is polynomial-time computable. Thus, there may be a possibility of using this or other relaxations to develop search heuristics for HTN planning.

Learning hierarchical task network domains from partially observed plan traces

Hierarchical Task Network (HTN) planning is an effective yet knowledge intensive problem-solving technique. It requires humans to encode knowledge in the form of methods and action models. Methods describe how to decompose tasks into subtasks and the preconditions under which those methods are applicable whereas action models describe how actions change the world. Encoding such knowledge is a difficult and time-consuming process, even for domain experts. In this paper, we propose a new learning algorithm, called HTNLearn, to help acquire HTN methods and action models. HTNLearn receives as input a collection of plan traces with partially annotated intermediate state information, and a set of annotated tasks that specify the conditions before and after the tasks' completion. In addition, plan traces are annotated with potentially empty partial decomposition trees that record the processes of decomposing tasks to subtasks. HTNLearn outputs are a collection of methods and action models. HTNLearn first encodes constraints about the methods and action models as a constraint satisfaction problem, and then solves the problem using a weighted MAX-SAT solver. HTNLearn can learn methods and action models simultaneously from partially observed plan traces (i.e., plan traces where the intermediate states are partially observable). We test HTNLearn in several HTN domains. The experimental results show that our algorithm HTNLearn is both effective and efficient.

A Research of HTN-Based Anytime Planning Algorithm on FPS Games

With the development of game industry, the requirements of game AI are also increasing. Recently a planning formalism called HTN (Hierarchical Task Network) was used to design AI behaviors. In this paper we describe how to implement an anytime planner based on HTN to handle a dynamic environment, which is the key feature of computer game worlds. This planner allows agent to interrupt its sequence of actions at anytime based on the changes of environment meanwhile return a valid replacement plan. We use Unreal Tournament 2004 which is a famous first-person shooter game as the platform and test our anytime planner.