Planning with Concurrency, Time and Resources

This chapter proposes to model a planning problem (e.g., the control of a satellite system) by identifying a set of relevant components in the domain (e.g., communication channels, on-board memory or batteries), which need to be controlled to obtain a desired temporal behavior. The domain model is enriched with the description of relevant constraints with respect to possible concurrency, temporal limits and scarce resource availability. The paper proposes a planning framework based on this view that relies on a formalization of the problem as a Constraint Satisfaction Problem (CSP) and defines an algorithmic template in which the integration of planning and scheduling is a fundamental feature. In addition, the paper describes the current implementation of a constraint-based planner called OMP that is grounded on these ideas and shows the role constraints have in this planner, both at domain description level and as a guide for problem solving.

A Framework for Hybrid and Analogical Planning

This chapter describes a model and an underlying theoretical framework for hybrid planning. Modern planning domain description languages are based on sentential representations. Sentential formalisms produce problem encodings that often lead the system to carry out large amounts of superfluous operations, causing a loss in performance. This chapter illustrates how techniques from the area of knowledge representation and reasoning (in particular, analogical representations) can be adopted to develop more efficient domain description languages. Although often more efficient, analogical representations are generally less expressive than sentential ones. A framework for planning with hybrid representations is thus proposed, in which sentential and analogical descriptions can be integrated and used interchangeably, thereby overcoming the limitations and exploiting the advantages of both paradigms.

Efficiently Dispatching Plans Encoded as Simple Temporal Problems

The Simple Temporal Problem (STP) formalism was developed to encode flexible quantitative temporal constraints, and it has been adopted as a commonly used framework for temporal plans. This chapter addresses the question of how to automatically dispatch a plan encoded as an STP, that is, how to determine when to perform its constituent actions so as to ensure that all of its temporal constraints are satisfied. After reviewing the theory of STPs and their use in encoding plans, we present detailed descriptions of the algorithms that have been developed to date in the literature on STP dispatch. We distinguish between off-line and online dispatch, and present both basic algorithms for dispatch and techniques for improving their efficiency in time-critical situations.

Continuous Planning for Virtual Environments

This chapter discusses the application of intelligent planning techniques to virtual agent environments as a mechanism to control and generate plausible virtual agent behaviour. The authors argue that the real world-like nature of intelligent virtual environments (IVEs) presents issues that cannot be tackled with a classic, off-line planner where planning takes place beforehand and execution is performed later, based on a set of precompiled instructions. What IVEs call for is continuous planning, a generative system that will work in parallel with execution, constantly re-evaluating world knowledge and adjusting plans according to new data. The authors argue further on the importance of incorporating the modelling of the agents’ physical, mental and emotional states as an inherent feature in a continuous planning system targeted towards IVEs, necessary to achieve plausibility in the produced plans and, consequently, in agent behaviour.

Coordination in Multi-Agent Planning with an Application in Logistics

Multi-agent planning comprises planning in an environment with multiple autonomous actors. Techniques for multi-agent planning differ from conventional planning in that planning activities are distributed and the planning autonomy of the agents must be respected. We focus on approaches to coordinate the multi-agent planning process. While usually coordination is intertwined with the planning process, we distinguish a number of separate phases in the planning process to get a clear view on the different role(s) of coordination. In particular, we discuss the pre-planning coordination phase and post-planning coordination phase. In the pre-planning part, we view coordination as the process of managing (sub) task dependencies and we discuss a method that ensures complete planning autonomy by introducing additional (intra-agent) dependencies. In the post-planning part, we will show how agents can improve their plans through the exchange of resources. We present a plan merging algorithm that uses these resources to reduce the costs of independently developed plans. This (any-time) algorithm runs in polynomial time.

Machine Learning for Adaptive Planning

This chapter is concerned with the enhancement of planning systems using techniques from Machine Learning in order to automatically configure their planning parameters according to the morphology of the problem in hand. It presents two different adaptive systems that set the planning parameters of a highly adjustable planner based on measurable characteristics of the problem instance. The planners have acquired their knowledge from a large data set produced by results from experiments on many problems from various domains. The first planner is a rule-based system that employs propositional rule learning to induce knowledge that suggests effective configuration of planning parameters based on the problem’s characteristics. The second planner employs instance-based learning in order to find problems with similar structure and adopt the planner configuration that has proved in the past to be effective on these problems. The validity of the two adaptive systems is assessed through experimental results that demonstrate the boost in performance in problems of both known and unknown domains. Comparative experimental results for the two planning systems are presented along with a discussion of their advantages and disadvantages.

Declarative Planning and Knowledge Representation in an Action Language

This chapter introduces planning and knowledge representation in the declarative action language K. Rooted in the area of Knowledge Representation & Reasoning, action languages like K allow the formalization of complex planning problems involving non-determinism and incomplete knowledge in a very flexible manner. By giving an overview of existing planning languages and comparing these against our language, we aim on further promoting the applicability and usefulness of high-level action languages in the area of planning. As opposed to previously existing languages for modeling actions and change, K adopts a logic programming view where fluents representing the epistemic state of an agent might be true, false or undefined in each state. We will show that this view of knowledge states can be fruitfully applied to several well-known planning domains from the literature as well as novel planning domains. Remarkably, K often allows to model problems more concisely than previous action languages. All the examples given can be tested in an available implementation, the DLVK planning system.

Plan Optimization by Plan Rewriting

Planning by Rewriting (PbR) is a paradigm for efficient high-quality planning that exploits declarative plan rewriting rules and efficient local search techniques to transform an easy-to-generate, but possibly suboptimal, initial plan into a high-quality plan. In addition to addressing planning efficiency and plan quality, PbR offers a new anytime planning algorithm. The plan rewriting rules can be either specified by a domain expert or automatically learned. We describe a learning approach based on comparing initial and optimal plans that produce rules competitive with manually specified ones. PbR is fully implemented and has been applied to several existing domains. The experimental results show that the PbR approach provides significant savings in planning effort while generating high-quality plans.

Constraint Satisfaction for Planning and Scheduling

As the current planning and scheduling technologies are coming together by assuming time and resource constraints in planning or by allowing introduction of new activities during scheduling, the role of constraint satisfaction as the bridging technology is increasing and so it is important for researchers in these areas to understand the underlying principles and techniques. The chapter introduces constraint satisfaction technology with emphasis on its applications in planning and scheduling. It gives a brief survey of constraint satisfaction in general, including a description of mainstream solving techniques, that is, constraint propagation combined with search. Then, it focuses on specific time and resource constraints and on search techniques and heuristics useful in planning and scheduling. Last but not least, the basic approaches to constraint modelling for planning and scheduling problems are presented.

AI Planning and Intelligent Agents

This chapter provides an overview of complementary research in the active research areas: AI planning technology and intelligent agents technology. It has been widely acknowledged that modern intelligent agents approaches should combine methodologies, techniques and architectures from many areas of computer science, cognitive science, operation research, cybernetics, and so forth. AI planning is an essential function of intelligence that is necessary in intelligent agents applications. This chapter presents the current state-of-the-art in the field of intelligent agents, focusing on the role of AI planning techniques. It sketches a typical classification of agents, agent theories and architectures from an AI planning perspective, it briefly introduces the reader to the basic issues of AI planning, and it presents different AI planning methodologies implemented in intelligent agents applications. The authors aim at stimulating research interest towards the integration of AI planning with intelligent agents.