Abstract: This article analyzes the current state of case-based plan adaptation research. Rather than perform an exhaustive literature review, we identify six dimensions to classify the various contributions in the field and to analyze current issues and trends. These dimensions are: the type of adaptation, the role of the case, the case content, the use of case merging, the representation formalism, and the computational complexity of the algorithm. Our analysis clarifies some common misconceptions about plan adaptation and proposes a set of future research directions. 1.

Abstract. Many challenges remain in the development of tactical planning systems that will enable automated, cooperative replanning of routes and mission assignments for multiple unmanned ground vehicles (UGVs) under changing environmental and tactical conditions. We have developed such a planning system that uses an evolutionary algorithm to assign waypoints and mission goals to multiple UGVs so that they jointly achieve a set of mission goals. Our evolutionary system applies domain-specific genetic operators, termed tactical advocates because they capture specific tactical behaviors, to make targeted improvements to plans. The plans are evaluated using a set of tactical critics that together comprise a multiobjective fitness function. Each critic evaluates a plan against criteria such as avoiding an enemy or meeting mission goals. Experimental results show that this approach produces highquality plans with the potential for real-time dynamic replanning. 1

One of the main bottlenecks in deploying casebased planning systems is authoring the case-base of plans. In this paper we will present a collection of algorithms that can be used to automatically learn plans from human demonstrations. Our algorithms are based on the basic idea of a plan dependency graph, which is a graph that captures the dependencies among actions in a plan. Such algorithms are implemented in a system called Darmok 2 (D2), a case-based planning system capable of general game playing with a focus on real-time strategy (RTS) games. We evaluate D2 with a collection of three different games with promising results. 1

Abstract. Automatic case elicitation (ACE) is a learning technique in which a case-based reasoning system acquires knowledge automatically from scratch through repeated real-time trial and error interaction with its environment without dependence on pre-coded domain knowledge. ACE represents an alternative to manually constructed case bases and domain specific techniques, and is generally applicable to any domain for which knowledge can be obtained from a series of observations of an environment (e.g., checkers or massively multiplayer games). A priority is placed on maintaining the flexibility necessary to learn new domains with only negligible manual configuration. We found during testing that the current approach to ACE with a reliance on experience and exploration, while quite capable in the domain of checkers, did not perform adequately in the exponentially more complex domain of chess. Our results suggest that experience alone, without the ability to adapt for case differences between new and prior cases, is insufficient in more complex domains. 1

We present a robotic dialogue system built on casebased reasoning. The system is capable of solving references and manage sub-dialogues in a dialogue with an operator in natural language. The approach to handle dialogue acts and physical acts in a unison manner together with the use of plans and subplans makes the system very flexible. This flexibility is used for learning purposes where the operator teaches the system a new word and the new knowledge can directly be integrated and used in the old plans. The learning from explanation capability makes the system adaptable to the operator's use of language and the domain it is currently operating in. The implementation of a case-based planner suggested in the paper will further increase the learning and adaptation degree.

ter verkrijging van de graad van doctor

Planning, the generation of a course of action to achieve a set of goals, is an important technique in the development of intelligent agents. Heretofore, planning has been largely considered as a one-shot problem. However, in practice, we are often dealing with situations in which an existing plan has to be adapted. Not only might we be facing a dynamic environment that requires a plan to be repaired, but it may also be that we recognise the new planning problem as being similar to one that we have solved before (i.e. case-based planning). This paper investigates a plan adaptation framework based on SAT-encodings of the planning problem. Compilation techniques have been very successfully applied to planning, as evidenced by their success in recent planning competitions. So far, however, such techniques have not been used for plan adaptation purposes. This paper explores whether it is feasible to modify the generated SAT instances such as to encode information that was extracted from the solution to the original planning problem.

This paper presents a CBP-BDI planning model which incorporates a novel artificial neural network. The CBP-BDI model, which is integrated within an agent, is the core of a Multiagent System that allows managing the security in industrial environments. The BDI model integrates within a CBP engine of reasoning that incorporates ANN-based techniques, and in this way it is possible to adapt past experiences to generate new plans. The proposed model uses Self-Organized Maps to calculate optimum routes for the security guards. Besides, some technologies of Ambient Intelligence such as RFID and Wi-Fi are used to develop the intelligent environment that has been tested and analyzed in this paper.

Contents lists available at ScienceDirect Artificial Intelligence www.elsevier.com/locate/artint

Abstract. Plan diversity has practical value in multiple planning domains, including travel planning, military planning and game planning. Existing methods for obtaining plan diversity fall under two categories: quantitative and qualitative. Quantitative plan diversity is domain-independent and does not require extensive knowledge-engineering effort, but can fail to reflect plan differences that are truly meaningful to users. Qualitative plan diversity is based on domain-specific characteristics which human experts might use to differentiate between plans, thus being able to produce results of greater practical value. However, the approach to qualitative plan diversity previously proposed in generative planning assumes the availability of a domain metatheory, hence requiring substantial knowledge engineering effort. We propose a case-based planning method for obtaining qualitative plan diversity through the use of distance metrics which incorporate domain-specific content. No additional domain theory is necessary, thus considerably reducing the knowledge-engineering effort. To our knowledge, this is the first time qualitative plan diversity is being explored in a case-based planning context.