Chris Unsworth

The temporal knapsack problem and its solution

This paper introduces a problem called the temporal knapsack problem, presents several algorithms for solving it, and compares their performance. The temporal knapsack problem is a generalisation of the knapsack problem and specialisation of the multidimensional (or multiconstraint) knapsack problem. It arises naturally in applications such as allocating communication bandwidth or CPUs in a multiprocessor to bids for the resources. The algorithms considered use and combine techniques from constraint programming, artificial intelligence and operations research.

A Constraint Programming Approach to the Hospitals / Residents Problem

An instance Iof the Hospitals / Residents problem (HR) involves a set of residents (graduating medical students) and a set of hospitals, where each hospital has a given capacity. The residents have preferences for the hospitals, as do hospitals for residents. A solution of Iis a stable matching, which is an assignment of residents to hospitals that respects the capacity conditions and preference lists in a precise way. In this paper we present constraint encodings for HR that give rise to important structural properties. We also present a computational study using both randomly-generated and real-world instances. We provide additional motivation for our models by indicating how side constraints can be added easily in order to solve hard variants of HR.

A specialised binary constraint for the stable marriage problem

We present a specialised binary constraint for the stable marriage problem. This constraint acts between a pair of integer variables where the domains of those variables represent preferences. Our constraint enforces stability and disallows bigamy. For a stable marriage instance with n men and women we require n2 of these constraints, and the complexity of enforcing arc-consistency is O(n3). Although this is non-optimal, empirical evidence suggests that in practical terms our encoding significantly outperforms the optimal encoding given in [7] in both space and time.

Towards the Verification of Pervasive Systems.

Pervasive systems, that is roughly speaking systems that can interact with their environment, are increasingly common. In such systems, there are many dimensions to assess: security and reliability, safety and liveness, real-time response, etc. So far modelling and formalizing attempts have been very piecemeal approaches. This paper describes our analysis of a pervasive case study (MATCH, a homecare application) and our proposal for formal (particularly verification) approaches. Our goal is to see to what extent current state of the art formal methods are capable of coping with the verification demand introduced by pervasive systems, and to point out their limitations.

Limited discrepancy search revisited

Harvey and Ginsbergs limited discrepancy search (LDS) is based on the assumption that costly heuristic mistakes are made early in the search process. Consequently, LDS repeatedly probes the state space, going against the heuristic (i.e., taking discrepancies) a specified number of times in all possible ways and attempts to take those discrepancies as early as possible. LDS was improved by Richard Korf, to become improved LDS (ILDS), but in doing so, discrepancies were taken as late as possible, going against the original assumption. Many subsequent algorithms have faithfully inherited Korfs interpretation of LDS, and take discrepancies late. This then raises the question: Should we take our discrepancies late or early? We repeat the original experiments performed by Harvey and Ginsberg and those by Korf in an attempt to answer this question. We also investigate the early stopping condition of the YIELDS algorithm, demonstrating that it is simple, elegant and efficient.

Tightly coupled verification of pervasive systems

We consider the problem of verifying context-aware, pervasive, interactive systems when the interaction involves both system configuration and system use. Verification of configurable systems is more tightly coupled to design when the verification process involves reasoning about configurable formal models. The approach is illustrated with a case study: using the model checker SPIN and a SAT solver to reason about a configurable model of an activity monitor from the MATCH homecare infrastructure. Parts of the models are generated automatically from actual log files.

An introduction to pervasive interface automata

Pervasive systems are often context-dependent, component based systems in which components expose interfaces and offer one or more services. These systems may evolve in unpredictable ways, often through component replacement. We present pervasive interface automata as a formalism for modelling components and their composition. Pervasive interface automata are based on the interface automata of Henzinger et al [3], with several significant differences. We expand their notion of input and output actions to combinations of input, output actions, and callable methods and method calls. Whereas interface automata have a refinement relation, we argue the crucial relation in pervasive systems is component replacement , which must include consideration of the services offered by a component and assumptions about the environment. We illustrate pervasive interface automata and component replacement with a small case study of a pervasive application for sports predictions.

Specialised constraints for stable matching problems

The stable marriage problem (SM) and the Hospital / Residents problem (HR) are both stable matching problems. They consist of two sets of objects that need to be matched to each other; in SM men to women, and in HR residents to hospitals. Each set of objects expresses a ranked preference for the objects in the other set, in the form of a preference list. The problem is then to find a matching of one set to the other such that the matching is stable. A matching is stable iff it contains no blocking pairs. A blocking pair in a matching M consists of two objects x and y one from each set(x = man and y = woman for SM or x = hospital and y = resident in HR), such that x and y are not matched in M and both x and y would rather be matched to each other than to there assignment in M. The author is supported by EPSRC. Software support was given by an ILOG SAs academic grant.

Is my configuration any good: checking usability in an interactive sensor-based activity monitor

We investigate formal analysis of two aspects of usability in a deployed interactive, configurable and context-aware system: an event-driven, sensor-based homecare activity monitor system. The system was not designed from formal requirements or specification: we model the system as it is in the context of an agile development process. Our aim was to determine if formal modelling and analysis can contribute to improving usability, and if so, which style of modelling is most suitable. The purpose of the analysis is to inform configurers about how to interact with the system, so the system is more usable for participants, and to guide future developments. We consider redundancies in configuration rules defined by carers and participants and the interaction modality of the output messages.Two approaches to modelling are considered: a deep embedding in which devices, sensors and rules are represented explicitly by data structures in the modelling language and non-determinism is employed to model all possible device and sensor states, and a shallow embedding in which the rules and device and sensor states are represented directly in propositional logic. The former requires a conventional machine and a model-checker for analysis, whereas the latter is implemented using a SAT solver directly on the activity monitor hardware. We draw conclusions about the role of formal models and reasoning in deployed systems and the need for clear semantics and ontologies for interaction modalities.