Lukáš Chrpa

On exploiting structures of classical planning problems: generalizing entanglements

Much progress has been made in the research and development of automated planning algorithms in recent years. Though incremental improvements in algorithm design are still desirable, complementary approaches such as problem reformulation are important in tackling the high computational complexity of planning. While machine learning and adaptive techniques have been usefully applied to automated planning, these advances are often tied to a particular planner or class of planners that are coded to exploit that learned knowledge. A promising research direction is in exploiting knowledge engineering techniques such as reformulating the planning domain and/or the planning problem to make the problem easier to solve for general, state-of-the-art planners. Learning (outer) entanglements is one such technique, where relations between planning operators and initial or goal atoms are learned, and used to reformulate a domain by removing unneeded operator instances. Here we generalize this approach significantly to cover relations between atoms and pairs of operators themselves, and develop a technique for producing inner entanglements. We present methods for detecting inner entanglements and for using them to do problem reformulation. We provide a theoretical treatment of the area, and an empirical evaluation of the methods using standard planning benchmarks and state-of-the-art planners.

On mixed-initiative planning and control for Autonomous underwater vehicles

Supervision and control of Autonomous underwater vehicles (AUVs) has traditionally been focused on an operator determining a priori the sequence of waypoints of a single vehicle for a mission. As AUVs become more ubiquitous as a scientific tool, we envision the need for controlling multiple vehicles which would impose less cognitive burden on the operator with a more abstract form of human-in-the-loop control. Such mixed-initiative methods in goal-oriented commanding are new for the oceanographic domain and we describe the motivations and preliminary experiments with multiple vehicles operating simultaneously in the water, using a shore-based automated planner.

Portfolio-based planning: State of the art, common practice and open challenges

In recent years the field of automated planning has significantly advanced and several powerful domain-independent planners have been developed. However, none of these systems clearly outperforms all the others in every known benchmark domain. This observation motivated the idea of configuring and exploiting a portfolio of planners to perform better than any individual planner: some recent planning systems based on this idea achieved significantly good results in experimental analysis and International Planning Competitions. Such results let us suppose that future challenges of the Automated Planning community will converge on designing different approaches for combining existing planning algorithms. This paper reviews existing techniques and provides an exhaustive guide to portfolio-based planning. In addition, the paper outlines open issues of existing approaches and highlights possible future evolution of these techniques.

KEWI: A Knowledge Engineering Tool for Modelling AI Planning Tasks

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.

Automated Planning for Urban Traffic Control: Strategic Vehicle Routing to Respect Air Quality Limitations

The global growth in urbanisation increases the demand for services including road transport infrastructure, presenting challenges in terms of mobility. These trends are occurring in the context of concerns around environmental issues of poor air quality and transport related carbon dioxide emissions. One out of several ways to help meet these challenges is in the intelligent routing of road traffic through congested urban areas. Our goal is to show the feasibility of using automated planning to perform this routing, taking into account a knowledge of vehicle types, vehicle emissions, route maps, air quality zones, etc. Specifically focusing on air quality concerns, in this paper we investigate the problem where the goals are to minimise overall vehicle delay while utilising network capacity fully, and respecting air quality limits. We introduce an automated planning approach for the routing of traffic to address these areas. The approach has been evaluated on micro-simulation models that use real-world data supplied by our industrial partner. Results show the feasibility of using AI planning technology to deliver efficient routes for vehicles that avoid the breaking of air quality limits, and that balance traffic flow through the network.

Towards application of automated planning in urban traffic control

Advanced urban traffic control systems are often based on feed-back algorithms. For instance, current traffic control systems often operate on the basis of adaptive green phases and flexible co-ordination in road (sub) networks based on measured traffic conditions. However, these approaches are still not very efficient during unforeseen situations such as road incidents when changes in traffic are requested in a short time interval. Therefore, we need self-managing systems that can plan and act effectively in order to restore an unexpected road traffic situations into the normal order. A significant step towards this is exploiting Automated Planning techniques which can reason about unforeseen situations in the road network and come up with plans (sequences of actions) achieving a desired traffic situation. In this paper, we introduce the problem of self-management of a road traffic network as a temporal planning problem in order to effectively navigate cars throughout a road network. We demonstrate the feasibility of such a concept and discuss our preliminary evaluation in order to identify strengths and weaknesses of our approach and point to some promising directions of future research.

The Fifth International Competition on Knowledge Engineering for Planning and Scheduling: Summary and Trends

We review the 2016 International Competition on Knowledge Engineering for Planning and Scheduling (ICKEPS), the fifth in a series of competitions started in 2005. ICKEPS series focuses on promoting the importance of knowledge engineering methods and tools for automated Planning and Scheduling systems.

Autonomic System Architecture: An Automated Planning Perspective

Control systems embodying artificial intelligence (AI) techniques tend to be “reactive” rather than “deliberative” in many application areas. There arises a need for systems that can sense, interpret and deliberate with their actions and goals to be achieved, taking into consideration continuous changes in state, required service level and environmental constraints. The requirement of such systems is that they can plan and act effectively after such deliberation, so that behaviourally they appear self-aware. In this paper, we focus on designing a generic architecture for autonomic systems which is inspired by the Human Autonomic Nervous System. Our architecture consists of four main components which are discussed in the context of the Urban Traffic Control Domain. We also highlight the role of AI planning in enabling self-management property of autonomic systems. We believe that creating a generic architecture that enables control systems to automatically reason with knowledge of their environment and their controls, in order to generate plans and schedules to manage themselves, would be a significant step forward in the field of autonomic systems.

A Monte-Carlo path planner for dynamic and partially observable environments

In this paper, we present a Monte-Carlo policy rollout technique (called MOCART-CGA) for path planning in dynamic and partially observable real-time environments such as Real-time Strategy games. The emphasis is put on fast action selection motivating the use of Monte-Carlo techniques in MOCART-CGA. Exploration of the space is guided by using corridors which direct simulations in the neighbourhood of the best found moves. MOCART-CGA limits how many times a particular state-action pair is explored to balance exploration of the neighbourhood of the state and exploitation of promising actions. MOCART-CGA is evaluated using four standard pathfinding benchmark maps, and over 1000 instances. The empirical results show that MOCART-CGA outperforms existing techniques, in terms of search time, in dynamic and partially observable environments. Experiments have also been performed in static (and partially observable) environments where MOCART-CGA still requires less time to search than its competitors, but typically finds lower quality plans.

Ontological support for modelling planning knowledge

This paper describes the conceptual model underlying the Knowledge Engineering Web Interface (KEWI) which primarily aims to be used for modelling planning tasks in a semi-formal framework. This 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, including function terms as object references. For operational use, KEWI automatically generates PDDL. Experiments show that the generated PDDL can be processed by a planner without incurring significant drawbacks.