David Wilkins

Domain-independent planning: representation and plan generation

Abstract A domain-independent planning program that supports both automatic and interactive generation of hierarchical, partially ordered plans is described. An improved formalism makes extensive use of constraints and resources to represent domains and actions more powerfully. The formalism also offers efficient methods for representing properties of objects that do not change over time, allows specification of the plan rationale (which includes scoping of conditions and appropriately relating different levels in the hierarchy), and provides the ability to express deductive rules for deducing the effects of actions. The implications of allowing parallel actions in a plan or problem solution are discussed, and new techniques for efficiently detecting and remedying harmful parallel interactions are presented. The most important of these techniques, reasoning about resources, is emphasized and explained. The system supports concurrent exploration of different branches in the search, making best-first search easy to implement.

Policy-Based Cognitive Radios

We present a new language for expressing policies that allow opportunistic spectrum access while not causing interference. CoRaL has expressive constructs for numerical constraints, supports efficient reasoning, and will be verifiable. The language is extensible so that unanticipated policy types can be encoded. We also describe a policy reasoner that reasons about CoRaL policies, and show how this reasoner can be used with various cognitive radios (in this case, an XG radio) to guarantee policy-specified behaviors while allowing spectrum sharing.

A Policy Engine for Spectrum Sharing

We argue for a policy-based approach to increase spectrum availability. To this extend, we briefly summarize a new language for expressing policies that allow opportunistic spectrum access. A Policy Reasoner that reasons about these policies can be used with cognitive radios to guarantee policy- specified behaviors while allowing spectrum sharing. We present our policy reasoner design and we evaluated the reasoner in a demonstration. We describe the policies used in that demonstration and the results of the evaluation.

CoRaL--Policy Language and Reasoning Techniques for Spectrum Policies

We present the cognitive radio (policy) language (CoRaL), a new language for expressing policies that govern the behavior of cognitive radios that opportunistically share spectrum. A Policy Reasoner validates radio transmissions to ensure that they are compliant with the spectrum policies. The Policy Reasoner also discovers spectrum sharing opportunities by deriving what requirements must be fulfilled for transmissions to be valid, i.e., in compliance with policies. A novel mix of reasoning techniques is required to implement such a reasoner.

Using knowledge to control tree searching

PARADISE (PAttern Recognition Applied to DIrecting SEarch) uses a knowledge-based analysis and little searching to find the correct move in chess middle game positions. PARADISEs search does not have a depth limit or any other artificial effort limit. This paper describes the methods used to constrain the search. The ideas of using different strategies to show that one move is best and using ranges to express the values of moves (first developed in Berliners B* search), are extended and clarified. PARADISE combines these ideas with the use of plans, a threshold, and various measures of possibility. Examples are presented, including one in which PARADISE uses an indirect strategy to prove that one move is best without finding the winning line (a first for a chess program).

Cognitive Radio Policy Language and Policy Engine

This chapter discusses the cognitive radio policy language (PL) and policy engine (PE). Various design considerations apply to any PL and PE for CRs and DSA. Different choices lead to different expressiveness of language and varying reasoning capabilities. It provides an overview of the requirements for the PL and PE that inform design. There are many advantages in using a policy-based approach to CRs. Deployment delays are drastically reduced because a policy-based architecture enables policies, policy reasoners, and radio devices to be accredited separately. Policies can be used to describe preferences and constraints on parameters, such as priority of traffic, security, quality of service, and probability of interference. Radio behavior can be changed in flexible ways at runtime by changing such policies. Finally, by having policies with clear, easily understood semantics, one can coordinate a variety of organizational entities. For example, regulators can specify admissible transmission behavior in a policy, or network managers with proper authority can activate, or deactivate, policies to flexibly control the network.

Using chess knowledge to reduce search

The current generation of computer chess programs select a move by exploring huge lookahead trees (millions of positions). Human masters, on the other hand, appear to use a knowledge-intensive approach to chess (see Chapter 2). They seem to have a huge number of stored “patterns,” and analyzing a position involves matching these patterns to suggest plans for attack or defense. This analysis is verified and possibly corrected by a small search of the game tree (tens of positions). Since the best humans still play better chess than the best programs, it is reasonable to explore computer programming strategies for using chess knowledge rather than extensive searching. This chapter describes a program named PARADISE (PAttern Recognition Applied to DIrecting SEarch) which uses this approach in an attempt to find the best move in tactically sharp middle game positions from the games of chess masters.

Evaluation of a delay-tolerant ICN architecture

Simulation/emulation is key for early testing, assessment, and scalability evaluation of networking solutions for mobile ad-hoc networks (MANETs). If the solution is highly configurable - such as ENCODERS, SRIs delay-tolerant information-centric networking (ICN) solution - this type of evaluation is crucial. For effective modeling of information flows, the test framework needs to: (1) allow repeatable execution of scenarios with different patterns of network traffic, operating in different mobility and network-usage contexts, (2) provide a rich simulated environment that can model virtually any network topology and mobility, with high-fidelity device models, and (3) support flexible large-scale simulation, with the option of using virtual machines that execute the same code that would be used on an actual device. We describe our evaluation framework and the results of using it to develop and evaluate ENCODERS.

Teambotica: a robotic framework for integrated teaming, tasking, networking, and control

Teambotica is a research environment for the exploration of theories, designs and implementations of team-based robotics. In developing Teambotica, we found that many of the simplifying assumptions that are often taken in both multiagent systems and behavior-based robotics had to be discarded. Central to our approach is a multilevel agent architecture which is adaptive along a number of dimensions and which is based on a vertically integrated design that spans a wide range of operations, from team-level reasoning to low-level control. The design addresses a number of pertinent issues: the proper mix of deliberation and action, flexible networking support including planning for communications, adaptive task level control, team-based monitoring, and an open systems modularity that takes form-factor considerations seriously. We also describe simulation tools for development and discuss several robotic teams that we have demonstrated.

Using patterns and plans in chess

The purpose of this research is to investigate the extent to which knowledge can replace and support search in selecting a chess move and to delineate the issues involved. This has been carried out by constructing a program, paradise (PAttern Recognition Applied to Directing SEarch), which finds the best move in tactically sharp middle game positions from the games of chess masters. It encodes a large body of knowledge in the form of production rules. The actions of the rules post concepts in the data base while the conditions match patterns in the chess position and data base. The program uses the knowledge base to discover plans during static analysis and to guide a small tree search which confirms that a particular plan is best. The search is “small” in the sense that the size of the search tree is of the same order of magnitude as a human master’s search tree (tens and hundreds of nodes, not thousands to hundreds of thousands as in many computer chess programs). Once a plan is formulated, it guides the tree search for several ply and expensive static analyses (needed to analyize a new position) are done infrequently. paradise avoids placing a depth limit on the search (or any other artificial effort limit). By using a global view of the search tree, information gathered during the search and the analysis provided by the knowledge base, the program produces enough terminations to force convergence of the search. paradise has found combinations as deep as 19 ply.