Intelligent agents designed to work in complex, dynamic environments such as e-commerce must respond robustly and flexibly to environmental and circumstantial changes, including the actions of other agents. An agent must have the capability to deliberate about appropriate courses of action, which may include reprioritising tasks--whether goals or associated plans--aborting or suspending tasks, or scheduling tasks in a particular order. In this article we study mechanisms to enable principled suspend, resuming, and aborting of goals and plans within a Belief-Desire-Intention (BDI) agent architecture. We give a formal and combined operational semantics for these actions in an abstract agent language (CAN), thus providing a general mechanism that can be incorporated into several BDI-based agent platforms. The abilities enabled by our semantics provides an agent designer greater flexibility to direct agent operation, offering a generic means to manage the status of goals. We demonstrate the reasoning abilities enabled on a document workflow scenario.

Often, such as in the presence of conicts, an agent must choose between multiple intentions. The level of completeness of the intentions can be a factor in this deliberation. We sketch a pragmatic but principled mechanism for quantifying the level of completeness of goals in a Belief-DesireIntention{like agent. Our approach leverages previous work on resource and eects summarization but we go beyond by accommodating both dynamic resource summaries and goal eects, while also allowing a non-binary quantication of goal completeness.

A fundamental feature of intelligent agents is their ability to deliberate over their goals. Operating in an environment that may change in unpredictable ways, an agent needs to regularly evaluate whether its current set of goals is the most appropriate set to pursue. The management of goals is thus a key aspect of an agent’s architecture. Focusing on BDI agents, we consider the various types of goals studied in the literature, including both achievement and maintenance goals. We develop a detailed description of goal states (such as whether goals have been suspended or not), and a comprehensive suite of operations that may be applied to goals (including dropping, aborting, suspending and resuming them). We provide an operational semantics corresponding to this detailed description in an abstract agent language (CAN), and demonstrate on a detailed real-life scenario. The three key contributions of our generic framework for goal states and transitions are (1) to encompass both goals of accomplishment and rich goals of monitoring, (2) to provide the first specification of abort and suspend for all the common goal types, and (3) to account for plan execution as well as the dynamics of subgoaling. Our semantics clarifies how an agent can manage its goals, based on the decisions that it chooses to make, and further provides a foundation for correctness verification of agent behaviour.

Often, such as in the presence of conflicts, an agent must choose between multiple intentions. The level of completeness of the intentions can be a factor in this deliberation. We sketch a pragmatic but principled mechanism for quantifying the level of completeness of goals in a Belief-Desire-Intention--like agent. Our approach leverages previous work on resource and effects summarization but we go beyond by accommodating both dynamic resource summaries and goal effects, while also allowing a non-binary quantification of goal completeness.

Given the current set of intentions an autonomous agent may have, intention selection is the agent's decision which intention it should focus on next. Often, in the presence of conflicts, the agent has to choose between multiple intentions. One factor that may play a role in this deliberation is the level of completeness of the intentions. To that end, this paper provides pragmatic but principled mechanisms for quantifying the level of completeness of goals in a BDI-style agent. Our approach leverages previous work on resource and effects summarization but we go beyond by accommodating both dynamic resource summaries and goal effects, while also allowing a non-binary quantification of goal completeness. We demonstrate the computational approach on an autonomous robot case study.

Personal assistant agents capable of proactively offering assistance can be more helpful to their users through their ability to perform tasks that otherwise would require user involvement. This article characterizes the properties desired of proactive behavior by a personal assistant agent in the realm of task management and develops an operational framework to implement such capabilities. We present an extended agent architectural model that features a meta-level layer charged with identifying potentially helpful actions and determining when it is appropriate to perform them. The reasoning that answers these questions draws on a theory of proactivity that describes user desires and a model of helpfulness. Operationally, assistance patterns represent a compiled form of this knowledge, instantiating metareasoning over the agent's beliefs about its user's activities as well as over world state. The resulting generic framework for proactive goal generation and deliberation has been implemented as part of a personal assistant agent in the computer desktop domain.

Learning by demonstration technology has long held the promise to empower nonprogrammers to customize and extend software. We describe the deployment of a learning by demonstration capability to support user creation of automated procedures in a collaborative planning environment that is used widely by the U.S. Army. This technology, which has been in operational use since the summer of 2010, has helped to reduce user work loads by automating repetitive and time-consuming tasks. The technology has also provided the unexpected benefit of enabling standardization of products and processes.

Learning by demonstration technology has long held the promise to empower non-programmers to customize and extend software. We describe the deployment of a learning by demonstration capability to support user creation of automated procedures in a collaborative planning environment that is used widely by the U.S. Army. This technology, which has been in operational use since the summer of 2010, has helped to reduce user workloads by automating repetitive and time-consuming tasks. The technology has also provided the unexpected benefit of enabling standardization of products and processes.

Deliberation over and management of goals is a key aspect of an agent's architecture. We consider the various types of goals studied in the literature, including performance, achievement, and maintenance goals. Focusing on BDI agents, we develop a detailed description of goal states (such as whether goals have been suspended or not) and a comprehensive suite of operations that may be applied to goals (including dropping, aborting, suspending and resuming them). We show how to specify an operational semantics corresponding to this detailed description in an abstract agent language (CAN). The three key contributions of our generic framework for goal states and transitions are (1) to encompass both goals of accomplishment and rich goals of monitoring, (2) to provide the first specification of abort and suspend for all the common goal types, and (3) to account for plan execution as well as the dynamics of sub-goaling.

The Adept Task Learning system is an end-user programming environment that combines programming by demonstration and direct manipulation to support customization by nonprogrammers. Previously, Adept enforced a rigid procedure-authoring workflow consisting of demonstration followed by editing. However, a series of system evaluations with end users revealed a desire for more feedback during learning and more flexibility in authoring. We present a new approach that interleaves incremental learning from demonstration and assisted editing to provide users with a more flexible procedure-authoring experience. The approach relies on maintaining a "soup" of alternative hypotheses during learning, propagating user edits through the soup, and suggesting repairs as needed. We discuss the learning and reasoning techniques that support the new approach and identify the unique interaction design challenges they raise, concluding with an evaluation plan to resolve the design challenges and complete the improved system.

Introduction. Deliberation over courses of action to pursue is fundamental to agent systems. Agents designed to work in dynamic environments, such as a rescue robot or an online travel agent, must be able to reason about what actions they should take, incorporating deliberation into their execution cycle, reviewing decisions and taking corrective action with appropriate focus and frequency. Not only must agents reason about the effects of their courses of action, they must also consider the semantics of these corrective actions. Systems based on the well-known Belief-Desire-Intention (BDI) framework most often ascribe a set of goals to the agent, which is equipped with various techniques to deliberate over and manage this set. The centrality of reasoning over goals is seen in the techniques investigated in the literature, which include subgoaling and plan selection, detection and resolution of conflicts or opportunities for cooperation [9], checking goal properties to specification [10, 5], failure recovery and planning [2], and dropping, aborting, or suspending and resuming goals [7]. A variety of goals are described in the literature, including goals of performance of a task, achievement of a state, querying truth of a statement, testing veracity of beliefs, and maintenance of a condition [1, 11]. An agent must manage such a variety of goals, while incorporating pertinent sources of information into its decisions over them, such as preferences, quality goals, motivational goals, and advice [10]. The complexity of agent goal management stems from this combination of the variety of goals and the breadth of deliberation considerations. It is furthered because each goal can be dropped, aborted, suspended, or resumed (as illustrated in Figure 1) at arbitrary times. While goals themselves are static (i.e., they are specified at design time, and do not change during execution), their behaviour is dynamic: a goal may undergo a variety of changes of state during its execution cycle [5]. This evolution may include its initial adoption by the agent, being actively pursued, being suspended and then later resumed, and eventually succeeding (or failing). (Maintenance goals have a subtle life-cycle: the goal is retained even when the desired property is true; it is possible that such goals are never dropped.) Our work analyzes the behaviour of the above types of goals, including the behaviour when goals are aborted or suspended. We consider the complete life-cycle of goals, from their initial adoption by the agent to the time when they are no longer of interest, and all stages in between; we account for the dynamics of plan execution and sub-goaling. We develop a generic framework for goal states and transitions that captures the life-cycle of goals—shown in summary in Figure 1; the Active and Suspended states decomposed further [8]—

The ability to proactively offer assistance promises to make personal agents more helpful to their users. We characterize the properties desired of proactive behaviour by a personal assistant agent in the realm of task management, and present an extended agent cognition model that features a meta-level layer charged with identifying potentially helpful actions and determining when it is appropriate to perform them. The reasoning that answers these questions draws on a theory of proactivity that describes user desires and a model of helpfulness. Operationally, assistance patterns represent a compiled form of this knowledge, instantiating meta-cognition over the agent's beliefs about its user's activities as well as over world state. We have implemented the resulting generic framework for proactive goal generation and deliberation as part of a personal assistant agent in the desktop domain.

Intelligent agents designed to work in complex, dynamic environments must respond robustly and flexibly to environmental and circumstantial changes. An agent must be capable of deliberating about appropriate courses of action, which may include reprioritising goals, aborting particular tasks, or scheduling tasks in a particular order. This paper investigates the incorporation of a mechanism to suspend and reconsider tasks within a BDI-style architecture. Such an ability provides an agent designer greater flexibility to direct agent operation, and it offers a generic means for handling conflicts between tasks. We investigate conditions under which a goal or a plan may be suspended, the process for suspending it, and the appropriate behaviours upon resumption. We give an operational semantics for suspending tasks in terms of the abstract agent language CAN, thus providing a general mechanism that can be incorporated into any BDI-based agent programming language.

We demonstrate an intelligent personal assistant agent that has been developed to aid a busy knowledge worker in managing time commitments and performing tasks. The PExA agent draws on a diverse set of AI technologies that are linked within the SPARK BDI agent framework. We focus on our agent's ability to provide assistance within the context of current user activities, based on its recognition of user workflows and their progress, and on its context-sensitive proactive suggestions. We have instrumented a common suite of desktop applications so that, endowed with a sophisticated workflow tracker, PExA has the ability to pervasively monitor the user's desktop activities. PExA follows and responds to the user's progress on shared tasks, and is highly user-centric in its support for user needs and its adaptivity to user working style and preferences.

Problem reduction is a basic concept behind much of AI search. The standard approach is to view problem reduction in terms of AND/OR graphs and then to concentrate on graph search techniques to nd a solution to the original problem. However these techniques depend upon global graph properties. By taking a more object-oriented approach we can ignore the graphs as a whole and instead need only deal with the local properties of individual problem reduction steps. This paper presents the GAO algorithm, which is a generalization of the AO algorithm based on this shift in viewpoint. As well as encompassing the behaviours of existing algorithms (such as A , AO , MINIMAX , , SSS , and SCOUT ), GAO goes further. It allows the seamless integration of di erent problem domains, di erent search strategies, and even non-search problem-solving techniques, within the same algorithm.

Intelligent agents that are intended to work in dynamic environments must be able to gracefully handle unsuccessful tasks and plans. In addition, such agents should be able to make rational decisions about an appropriate course of action, which may include aborting a task or plan, either as a result of the agent's own deliberations, or potentially at the request of another agent. In this paper we investigate the incorporation of aborts into a BDI-style architecture. We discuss some conditions under which aborting a task or plan is appropriate, and how to determine the consequences of such a decision. We augment each plan with an optional abort-method, analogous to the failure method found in some agent programming languages. We provide an operational semantics for the execution cycle in the presence of aborts in the abstract agent language CAN, which enables us to specify a BDI-based execution model without limiting our attention to a particular agent system (such as JACK, Jadex, Jason, or SPARK). A key technical challenge we address is the presence of parallel execution threads and of sub-tasks, which require the agent to ensure that the abort methods for each plan are carried out in an appropriate sequence.

Intelligent agents that are intended to work in dynamic environments must be able to gracefully handle unsuccessful goals and plans. In addition, such agents should be able to make rational decisions about an appropriate course of action, which may include aborting a goal or plan, perhaps at the request of another agent, or as a result of the agent’s own deliberations. In this paper we investigate the incorporation of aborts into a BDI-style architecture. We discuss some conditions under which aborting a goal or plan is appropriate, and how to determine the consequences of such a decision. We augment each plan with an optional abort-method, analogous to the failure method found in some agent programming languages. We provide an operational semantics for the execution cycle in the presence of aborts in the abstract agent language CAN, which enables us to specify a BDI-based execution model without limiting our attention to a particular agent system (such as JACK, Jadex, Jason, or SPARK). A key technical challenge we address is the presence of parallel execution threads and of sub-goals, which require the agent to ensure that the abort methods for each plan are carried out in an appropriate sequence.

We describe an intelligent personal assistant that has been developed to aid a busy knowledge worker in managing time commitments and performing tasks. The design of the system was motivated by the complementary objectives of (1) relieving the user of routine tasks, thus allowing her to focus on tasks that critically require human problem-solving skills, and (2) intervening in situations where cognitive overload leads to oversights or mistakes by the user. The system draws on a diverse set of AI technologies that are linked within a Belief-Desire-Intention (BDI) agent system. Although the system provides a number of automated functions, the overall framework is highly user centric in its support for human needs, responsiveness to human inputs, and adaptivity to user working style and preferences.

The A search algorithm nds an optimal path through a search graph. It is easily modi ed to return multiple paths in decreasing order of optimality, but it cannot handle conjunctive problems. The AO search algorithm does handle conjunctive problems and nds an optimal subgraph of an acyclic AND/OR graph. However, it returns only one solution and can only handle totally independent conjunctive problems. This paper describes how a combination of the two algorithms (the AAO algorithm) can be used to return multiple subgraphs of an acyclic AND/OR graph in order of optimality. In addition, it can handle dependencies between conjuncts, but at increased computational cost.