A fundamental assumption made by classical AI planners is that there is no uncertainty in the world: the planner has full knowledge of the conditions under which the plan will be executed and the outcome of every action is fully predictable. These planners cannot therefore construct contingency plans, i.e., plans in which different actions are performed in different circumstances. In this paper we discuss some issues that arise in the representation and construction of contingency plans and describe Cassandra, a partial-order contingency planner. Cassandra uses explicit decision-steps that enable the agent executing the plan to decide which plan branch to follow. The decision-steps in a plan result in subgoals to acquire knowledge, which are planned for in the same way as any other subgoals. Cassandra thus distinguishes the process of gathering information from the process of making decisions. The explicit representation of decisions in Cassandra allows a coherent approach to...

O-Plan is a command, planning and control architecture with an open modular structure intended to allow experimentation on, or replacement of, various components. The research is seeking to determine which functions are generally required in a number of application areas and across a number of different command, planning, scheduling and control systems. O-Plan aims to demonstrate how a planner, situated in a task assignment and plan execution (command and control) environment, and using extensive domain knowledge, can allow for flexible, distributed, collaborative, and mixedinitiative planning. The research is seeking to verify this total systems approach by studying a simplified three-level model with separable task assignment, plan generation and plan execution agents. O-Plan has been applied to logistics tasks that require flexible response in changing situations. Summary The O-Plan research and development project is seeking to identify re-usable modules and interfaces within plan...

Condition satisfaction in planning has received a great deal of experimental and formal attention. A "Truth Criterion" lies at the heart of many planners and is critical to their capabilities and performance. However, there has been little study of ways in which the search space of a planner incorporating such a Truth Criterion can be guided. The aim of this document is to give a description of the use of condition "type" information to inform the search of an AI planner and to guide the production of answers by a planner's truth criterion algorithm. The authors aim to promote discussion on the merits or otherwise of using such domain-dependent condition type restrictions as a means to communicate valuable information from the domain writer to a general purpose domain-independent planner 1 . Introduction to Condition Typing Research in AI planning has introduced a range of progressively more powerful techniques to address increasingly more realistic applications (Allen, Hendler & Ta...

The development of ambient intelligence (AmI) applications that effectively adapt to the needs of the users and environments requires, among other things, the presence of planning mechanisms for goal-oriented behavior. Planning is intended as the ability of an AmI system to build a course of actions that, when carried out by the devices in the environment, achieve a given goal. The problem of planning in AmI has not been yet adequately explored in literature. In this paper we propose a planning system for AmI applications, based on the Hierarchical Task Network (HTN) approach and called D-HTN, able to find courses of actions to address given goals. The plans produced by D-HTN are flexibly tailored to exploit in the best way the capabilities of the devices currently available in the environment. We discuss both the architecture and the implementation of D-HTN. Moreover, we present some of the experimental results that validated the proposed planner in a realistic application scenario in which an AmI system monitors and answers the needs of a diabetic patient.

In this article we review knowledge representation formalisms that lend themselves to the representation of capabilities of intelligent agents. The aim of representing capabilities is, of course, that we want to reason about them. The reasoning task we are most interested in is capability brokering, i.e. the task of finding an agent which has a capability that can be used to address a given problem. Thus, the first area we review here is agent cooperation and communication from which the problem originates.

In this thesis we investigate the relationships between the symbolic and sub-symbolic methods used for controlling agents by artificial intelligence, focusing in particular on methods that learn. In light of the strengths and weaknesses of each approach, we propose a hybridisation of symbolic and subsymbolic methods to capitalise on the best features of each. We implement such a hybrid system, called Rachel which incorporates techniques from Teleo-Reactive Planning, Hierarchical Reinforcement Learning and Inductive Logic Programming. Rachel uses a novel representation of be-haviours, Reinforcement-Learnt Teleo-operators (RL-Tops), which defines the behaviour in terms of its desired consequences but leaves the implementation of the policy to be learnt by reinforcement learning. An RL-Top is an abstract, symbolic description of the purpose of a behaviour, and is used by Rachel both as a planning operator and as the definition of a reward function by which the behaviour can be learnt. Two new

This working paper has been prompted by recent work on hierarchic nonlinear planners (NONLIN is such a system) and on some work in temporal logics to support planning. The only easily available reference for NONLIN was an IJCAI-77 (Tate, 1977) paper which concentrated on one particular feature. The complete description is available in an Edinburgh Department of Artificial Intelligence Research Report (Tate, 1975). The NONLIN features to be highlighted below will be related to the different terminology used by other workers and I will show where their systems have gone beyond the applicability of NONLIN. All of the work on hierarchic non-linear planners has a root in Sacerdoti's landmark work on the NOAH planner at SRI (Sacerdoti,