Automated Planning for Automated Systems

A new Masters by Research (MRes) degree in Automated Planning for Autonomous Systems is now available to students worldwide from the Department of Computer and Information Sciences at the University of Strathclyde. The university has, for many years enjoyed an international reputation for its excellence in research. This unique Masters programme draws on the department's leading expertise within this field.

Automated Planning for Autonomous Systems

The Masters by Research (MRes) degree in Automated Planning for Autonomous Systems is an exciting opportunity to be at the forefront of a technology that will play a pivotal role in the next generation of automated systems. Areas such as assistive technologies, robotics, survey systems, security systems, condition monitoring and supply systems all require autonomous planning, control and execution layers to exploit their potential.

The course is a one year, intensive Masters course. It consists of a taught component that covers the most recent advances in the field, followed by a project designed to offer direct experience in the deployment of AI Planning and autonomous systems. The course has been designed with advice from industries representing current or future users of this technology and it will provide a basis for those who want to act as technical ambassadors, technology leaders or researchers in this field.

Autonomous Intelligent Control

Technological advances are providing us with more and more sophisticated systems - robots, fly-by-wire aircraft, telecommunication networks, etc. The drive is to make our environment more efficient, more convenient and more capable. But effective control of these systems is an increasing challenge.

A key supporting technology in managing these complex systems is autonomous intelligent control. Autonomous control relies on intelligent software to make decisions, in which a key element is the ability to plan. Planning is the problem-solving heart in managing resources, time and energy, to achieve the objectives of an autonomous controller for a complex system.

Planning Versus Reactive Control

Why do we need planning to control a system? After all, many systems are controlled perfectly successfully using tightly-coupled simple reactive systems.

Reactive control is inadequate when decisions made about current behaviour constrain future choices. For example, the autonomous underwater vehicle, Autosub, intended for under-ice exploration, was lost in a mission in 2005: the failure of its autonomous reactive control systems to recognise that it was attempting to surface while still under-ice was partly responsible for this loss and illustrates how important predictive control can be when making decisions about irreversible actions. If a current decision can cause consumption of a limited resource preventing action later, then an intelligent choice must be made between immediate reward and deferred reward. The satellite, Earth-Observing-1, is an example of a successful deployment of intelligent control software, including planning, using predictive control and intelligent decision-making to control a remote system.

Planning is required when the control loop that links the decision-maker to execution contains significant delay compared with execution time of actions. Decisions about actions must then be made before they are executed, because reactive control is impossible. Remote space probes represent one class of devices for which this situation holds. Making decisions about actions before they are executed is the central role of planning. Operating in hazardous conditions where communications are limited, including disaster recovery situations, nuclear waste management and condition inspection in areas with very limited access are all examples of problems in which planning can play a central role.

Planning Technologies

Planning research has made dramatic leaps, with modern systems tackling far more complex and elaborate problems than previously possible. These advances make it possible to deploy planning technology within a new generation of intelligent systems, harnessing their power and improving exploitation. Planning is part of Artificial Intelligence and so draws on fundamental technologies relevant to AI problem-solving. Modelling, constraint reasoning, search techniques, heuristics and abstraction are standard tools for the AI scientist and crucial to modern planning.

The modelling problem in planning involves deciding on levels of abstraction, capturing time and resource constraints and considering the impact of uncertainty. The focus of much research in the field is on how to handle the complex interactions that arise between these elements.

Constraint reasoning, heuristic search and abstraction all play a role in planning. Modern systems combine these elements in powerful strategies for problem-solving. Applications of planning systems harness these techniques in different ways and a challenge is to find ways to configure these combinations, both manually and automatically.

Planning in Context

Planners are only part of the technology required for intelligent control of autonomous systems. To exploit them requires a framework for execution, execution monitoring, diagnosis and plan repair. This is illustrated in application systems, ranging from fully autonomous systems such as the Earth-Observing-1 system, to human-in-the-loop systems such as the SIADEX forest fire-fighting planning system. Systems using planners include virtual systems such as web-services planners, humans, tourist trip planners and robots. There are many examples of planning applications, and of problems that can be solved by planners, in all walks of life. The new MRes course places planning in the context of applications. The course is exploiting partnerships both inside and outside the university, to give opportunities for building autonomous control systems in exciting new areas of application.

Choose To Lead In Advanced Robotics

Students who pursue this course will gain world-leading experience of the practical development of real autonomous intelligent systems for the future. This will put them in an ideal position to contribute to developments in advanced robotics as the importance of autonomy in all areas of modern life continues to grow into the future.

Automated Planning for Autonomous System Application

Anyone interested in pursuing this opportunity should download the postgraduate application form at: www.strath.ac.uk/prospectus/postgraduateapplications,

Or apply online at: http://applicants.strath.ac.uk.

Further information about the course and funding opportunities can be obtained from the Departmental web page at: http://www.cis.strath.ac.uk.