Automatic Planning and Decision-Making
Abstract: The field of AI planning has seen a recent rise in interest as more realistic domains have been tackled by automatic planning systems. In particular, planning is now used to control the daily activities of the NASA Mars rovers, make high-level decisions for DARPA Grand Challenge vehicles, and in many other real-world applications. In this series of lectures I will cover the background of AI planning techniques and lead up to an examination of the state of the art in both classical planning domains and planning under uncertainty. The lectures will be motivated by understanding the Mars rovers planning problem, how the current rover planning system works, what future challenges exist, and how these problems are being overcome in the latest planning systems.
XML and Web Application Programming
Programming with XML and WWW leads to many interesting opportunities and challenges to design of programming languages and static analyses. These lectures will first present an overview of the basic XML technologies, focusing on formal models of schema languages. We then study the state of the art of programming languages designed for processing XML data and discuss the challenges for further research. Finally, we survey the key ideas in modern frameworks for Web application programming.
See recommendation for additional reading.
Interactive Computation: A New Paradigm of Problem Solving
These talks will explore the transition of computer science from algorithmic Turing machine models to models of interactive problem solving.
Computing has advanced from the time-independent transformation of functions from inputs to outputs to the time-dependent performance of tasks and services over interactive environmental computation sequences. Moreover, scientific models of mathematics and physics may be as flawed as models of computation, since their validity may depend on inappropriate decision making rather than on legitimate modes of reasoning and theorem proving. I shall review my 15 years of work on interactive computing, including a recent book co-edited with my colleagues, and relate interaction to the work of Alan Turing, Kurt Godel and other researchers who question the earlier mathematical model of Hilbert on the provability of theorems. It is shown that the contradiction between the "Strong Church-Turing thesis" of the 1960s and the original Church-Turing thesis of the 1930s has resulted in a flawed model of computation that our interactive model is designed to correct.
  • Johannes Helander
Microsoft Research, Redmond.
Service Oriented Embedded Computing
Consumer electronics and other embedded devices can be more useful if they can take advantage of each other, more powerful computers, and the Internet. Unfortunately making systems distributed also make them more complicated and thus harder to program and harder to use. Thinking of the device capabilities and even the physical world that the devices interact with as services may aid in making distributed systems more manageable. A service can run on various types of hardware and on different generations of devices. A user can access the functions without having to know what (if any) software is on the device. To make service oriented embedded computing a reality we need methodologies for constructing, associating, managing, and programming services. We need a common way for representing data, common protocols, and ways to efficiently adapt software to new platforms. We need to make it easy for humans to control these devices. Finally, since these devices interact with the real world, they also need to be reliable and work in real-time.
The lectures will examine some techniques for making service oriented embedded computing work. Topics include
- XML in embedded systems. Examples of Web Services.
- Transformation of XML to runtime data structures and integration with embedded programs.
- Scaling programs to different levels of parallelism, memory, and flexibility.
- Data centric vs. procedural services. Creating new services.
- Distributed real-time, stochastic planning.
- Trust, human interaction, association.