Search

Hedlamp is a UK EPSRC grant funded research project in which we aim to tackle challenges with knowledge engineering of automated planning techniques when applied to real applications. Normally, successful deployment of planning technology relies on groups of planning experts encoding detailed domain models and investing large amounts of time maintaining them.We are developing a high level, application-oriented knowledge engineering framework usable by application developers who want to experiment with the potential of AI Planning,while encoding a precise domain model of some valuable application area. We are developing tools and theory for the framework which support knowledge acquisition, validation and operationality of the domain model. In particular, this project aims to explore the opportunities of applying model translation, adaptation and reformulation techniques to improve the model’s quality, and that of the planning function of which it is a part.In this paper we outline the main areas that Hedlamp has addressed,and overview an automated knowledge acquisition technique that has been tested with real industrial process data.

This paper introduces the Knowledge Engineering Web Interface (KEWI) which primarily aims to be used for modelling automated planning tasks in a semi-formal framework. The conceptual model used to represent the declarative and procedural knowledge in KEWI is described formally. The model consists of three layers: a rich ontology, a model of basic actions, and more complex methods. It is this structured conceptual model based on the rich ontology that facilitates knowledge engineering. The focus of this paper is to show how the central knowledge model used in KEWI differs from a model directly encoded in PDDL, the language accepted by most existing planning engines. Specifically, the rich ontology enables a more concise and natural style of representation. For operational use, KEWI automatically generates PDDL. Initial experiments show that the generated PDDL can be processed by a planner without incurring significant drawbacks.

This paper reports on progress towards a tool for the representation of shared, procedural and declarative knowledge whose aim is to be used for various functions to do with the automation of a complex process control application - primarily to guide the response phase during an emergency situation, but also for supporting normal automated operation. The tool is a Knowledge Engineering Web Interface called KEWI. The focus of the paper is on the conceptual model used to represent the declarative and procedural knowledge. The model consists of three layers: an ontology, a model of basic actions, and more complex methods. It is this structured conceptual model that facilitates knowledge engineering. We are aiming to evaluate the use of a central knowledge model for a range of planning-related functions, where the parts of the model are automatically assembled e.g. into PDDL for operational use.

This paper describes a method for analyzing STRIPS-like planning domains by identifying static graphs that are implicit in the set of operators defining a planning domain. A graph consisting of nodes and possibly directed edges is acommon way to construct representations for many problems, including computational problems and problems of reasoning about action. Furthermore, there may be objects or properties related to the nodes of such a graph that may be modified by the operators in ways restricted by the graph. The formaldefinition of shift operators over static graphs as a domain feature is the main contribution of this paper. Such an analysis could be used for verification and validation of the planning domain when it is presented to the domain author who may or may not agree with the result of the analysis. The method described relies on domain features that can also be extracted automatically, and it works on domains rather than problems, which means the result is problem-independent. However, if problems are given further analysis may be perfomed. The method has been evaluated using a small number of planning domains drawn from the international planning competition.

This paper describes the OpenVCE system, which is an open-source environment that integrates Web 2.0 technology and a 3D virtual world space to support collaborative work, specifically in large-scale emergency response scenarios, where the system has been evaluated. The support is achieved through procedural knowledge that is available to the system. OpenVCE supports the distributed knowledge engineering of procedural knowledge in a semi-formal framework based on a wiki. For the formal aspect it relies on a representation used in AI planning, specifically, Hierarchical Task Networks, which corresponds naturally to the way emergency response procedures are described in existing Standard Operating Procedures. Knowledge engineering is supported by domain analysis that may highlight issues with the representation. The main contribution of this paper lies in a reasonably informal description of the analysis.The procedural knowledge available to OpenVCE can be utilized in the environment through plans generated by a planner and given to the users as intelligent, distributed to-do lists. The system has been evaluated in experiments using emergency response experts, and it was shown that procedural uncertainty could be improved, despite the complex and new technologies involved. Furthermore, the support for knowledge engineering through domain analysis has been evaluated using several domains from the International Planning Competition, and it was possible to bring out some issues with these examples.

We are working on support for distributed collaboration for the Dismounted Incident Collaboration Environment (DICE). The overall DICE project at the US Army Research Lab’s Human Research & Engineering Directorate is seeking to give support to offer best advice and specialist assistance when soldiers are injured and are being treated. Distributed collaboration, as required for the DICE community, is typical of the needs of many communities that work across national and organisational boundaries as in multi-national coalitions. This paper gives a summary of work in progress on the DICE project and its distributed collaboration experimentation.

This paper describes a prototype implementation of a configuration system which uses automated planning techniques to compute workflows between declarative states. The resulting workflows are executed using the popular combination of ControlTier and Puppet. This al- lows the tool to be used in unattended “autonomic” situ- ations where manual workflow specification is not feasi- ble. It also ensures that critical operational constraints are maintained throughout the execution of the work- flow. We describe the background to the configuration and planning techniques, the architecture of the proto- type, and show how the system deals with several exam- ples of typical reconfiguration problems.

This paper concerns the use of new media technologies, including virtual worlds and web 2.0, for on-line collaborative activities, and specifically for the provision of expert advice about the response to large-scale crises. Internet technologies in general offer rich possibilities for interactions involving remote experts; however, the diversity, novelty and power of these technologies are such that to introduce them into problem-solving episodes without first developing a model of the nature of those episodes and the type of collaborative support they require, risks confusing and discouraging users. After a brief discussion of the nature of distributed collaboration and the implications this has for any technical support, we describe a virtual collaboration environment that has been developed to foster task-focused communities and support them through specific problem-solving episodes, and present some of the results of evaluation experiments.

This paper defines a number of features that can be used to characterize planning domains, namely domain types, relation fluency, inconsistent effects and reversible actions. These features can be used to provide additional information about the operators defined in a STRIPS-like planning domain. Furthermore, the values of these features may be extracted automatically; efficient algorithms for this are described in this paper. Alternatively, where these values are specified explicitly by the domain author, the extracted values can be used to validate the consistency of the domain, thus supporting the knowledge engineering process. This approach has been evaluated using a number of planning domains, mostly drawn from the international planning competition. The results show that the features provide useful information, and can highlight problems with the manual formalization of planning domains.

This paper describes a system that supports the distributed development and deployment of Standard Operating Procedures.the system is based on popular, open-source wiki software for the SOP development, and the I-X task-centric agent framework for deployment. A preliminary evaluation using an SOP for virtual collaboration is described and shows the potential of the approach.

This paper describes some of the lessons learned from the FireGrid project. It starts with a brief overview of the project. The discussion of the lessons learned that follows is intended for others attempting to develop a similar system, where sensor data is used to steer a super-real time simulation in order to generate predictions that will provide decision support for emergency responders.

One of the problems that must be solved during coalition operations is the planning problem: finding a course of actions for the operation. Knowledge-based planning is a technique that can be used to address this problem. However, traditional planning has focussed on the finding of a plan that achieves a given goal or accomplishes a given task. During a coalition operation such a plan may need to be agreed between the different partiesinvolved in the coalition. In this paper we shall describe an approach that can be used for finding an agreed plan. This plan will include a set of decommitmentrules allowing participants to specify conditions under which they are allowed to deviate from the agreed course of action. The proposed algorithm can be used as the basis for an automated planner, but it should also be usable in a mixedinitiative setting.

We present an approach to distributed planning and coordination architecture for dynamic non-deterministic multi-actor mixed-initiative environment. The system provides exible planning, replanning, and task allocation. The key idea of the presented approach is in integration of (i) I-X hierarchical planner with (ii) agent-based architecture and (iii) commitment-based plan representation. The imple-mentation of the system was veried and evaluated on simulated environment. The experimental validation conrms the performance, stability, and robustness of the system in complex scenarios.

Qualitatively dierent plan generation is one aspect of planners that is usefulto execution agents that operate under uncertainty, such as those in coalition operations. The purpose of this research is to improve plan construction and selection. To do so, we dene the statistics that make plan evaluations most useful to planning and task assignment agents. Furthermore, we develop a method of optimizing and visualizing plans based on evaluation criteria. Experiments show that guiding the planner's search strategy based on plan evaluation criteria improves the diversity of qualitatively dierent plans.

This paper describes some of the lessons learned from the FireGrid project. It starts with a brief overview of the project. The discussion of the lessons learned that follows is intended for others attempting to develop a similar system, where sensor data is used to steer a super-real time simulation in order to generate predictions that will provide decision support for emergency responders.

The aim of this paper is to describe and solve a new type of planning problem in which a group of peers need to find and agree upon a plan that solves a given problem. Classical planning only addresses the finding of a plan, whereas agentresearch defines the concept of commitment which we will use to define the “agreed” aspect of a plan. The problem will be solved by introducing a level of indirection.

In this paper we describe the approach adopted and experiences gained during a project to develop a general architecture that aims to harness advanced sensor, modelling and Grid technologies to assist emergency responders in tackling emergencies (specifically fire emergencies). Here we focus on the command and control aspects of this architecture, and in particular, on a query-based approach that has been adopted to allow end users to interact with available models of physical and other phenomena. The development of this has provided a number of insights about the use of such models, which along with the approach itself, should be of interest to any considering similar applications.

The Collaborative Operations for Personnel Recovery (Co-OPR) project sought to provide collaborative task support for a Search and Rescue coordination center. The project aimed to create a prototype “Personnel Recovery (Experimental) Pack” (PREP) and to demonstrate its use. A number of requirements capture, knowledge gathering and transition workshops and meetings were held. This included an initial requirements setting workshop early in 2005, meetings at the USJFCOM Joint Personnel Recovery Agency’s (JPRA) Personnel Recovery Education and Training Center (PRETC) in Fredericksburg, Virginia in June 2005, a review meeting in Edinburgh in August 2005, and attendance at a Command Post Exercise (CPX) at the PRETC in November 2005. These initially established the potential areas for use of Co-OPR and I-X tools in training exercises. In the second project year such tools were developed and tested using a training exercise as held at the PRETC, using observers from PRETC and USJFCOM and an evaluation expert from USJFCOM/J9. The project was provided with a rich set of urban and rural scenarios by JPRA/PRETC which together are unclassified versions of scenarios used within the PRETC training courses and Command Post Exercises. At the time, these stretched the capabilities of the current and envisaged technologies within Co-OPR/I-X. Refinement of the scenarios alongside PRETC, and knowledge engineering to capture information on standard operating procedures and responses were a key part of making the work relevant to the potential real use of Co-OPR/I-X for Personnel Recovery. The core I-X technology was packaged into a number of checkpoint releases to make available the features required to meet the application needs. I-X version 4.3 released in November 2005 to checkpoint the results achieved on the first 12 months of work with the PRETC. It formed a basis for the work on really using the technology at the PRETC. New “white cell” aids for training were made available in an initial version. A new I-Sim simulation capability, and advanced option exploration tools have all been improved significantly to make them more usable, including the features of the I-Plan AI planner and its capability for plan repair after failures. The features of the Domain Editor (I-DE) and its ability to browse and update or augment standard operating procedure knowledge dynamically during missions were enhanced. The final release of I-X that includes all the new developments achieved in the Co-OPR project is version 4.5. The results of the work were packaged, along with Personnel Recovery domain specific models, as a web site and/or CD which could be considered as a prototype “Personnel Recovery (Experimental) Pack” of tools to assist a Joint Personnel Recovery Center (JPRC) and associated operational staff in performing their operations. The versions of PREP produced were used in one workshop or Command Post Exercise at the PRETC under guidance from Dr. Jeff Hansberger at training related workshops already organized by USJFCOM/J9 Expt. and Fred Kleibacker, the (now former) Director of the PRETC in Fredericksburg, Virginia. Co-OPR team members were engaged with these workshops to to show the tools in realistic settings, to assist with training where possible, and to gather experimental feedback. Realistic use of tools for Personnel Recovery requires that the systems can work with emerging technology for geo-positioning, survival radios, evasion aids, robotic or semiautomated rescue aids or robots, and doctrine or tactics, techniques and procedures for Personnel Recovery. A number of short studies of these “complementary technologies” were made which explored deployment and inter-working aspects of these with the Co-OPR/I-X technology.

The aim of this paper is to show how the rationale behind a plan can be recorded in the plan itself. The model which underlies the I-X framework will be described in detail, focussing on annotations. It is there that a planner can record the justifications for including components into the plan. Recording rationale information of this type can be used for a number of purposes in the life cycle of a plan, including plan indexing and retrieval, failure recovery, plan explanation and establishing trust as explained in this paper.

The aim of this paper is to describe the I-X system with its underlying representation: . The latter can be seen as a description of an agent’s intentions, which can be shared and communicated amongst multiple I-X agents to coordinate activities in an emergency response scenario. In general, an object describes the product of a synthesis task. In themulti-agent context it can be used to describe the intentions of an agent, although it also includes elements of beliefs about the world and goals to be achieved, thus showing a close relationship with the BDI agent model which we will explore in this paper. From a user’s perspective, I-X Process Panels can be used as an intelligent to-do list that assists emergency responders in applyingpre-defined standard operating procedures in different types of emergencies. In particular, multiple instances of the I-X Process Panels can be used as a distributed system to coordinate the efforts of independent emergency responders as well as responders within the same organization. Furthermore, it can be used as an agent wrapper for other software systems such as webservices to integrate these into the emergency response team asvirtual members. At the heart of I-X is a Hierarchical Task Network (HTN) planner that can be used to synthesize courses of action automatically or explore alternative options manually.