Determining the sequence of genes along a region of DNA from the results of experimental data is a difficult task called Map Assembly. A map indicates the order of the genes and other markers called restriction enzymes. It is a time consuming activity, carried out manually by the geneticist. The data from which maps are produced contain a high degree of error, due to experimental limitations, and several feasible solutions may be constructed from the same data. Distinguishing between competing solutions relies on the geneticist's subjective judgement. Although computer applications have been developed for map assembly they have been either restricted in the amount of data that could be handled or they addressed related problems. This thesis has investigated and developed suitable computer techniques for automating map assembly. A novel objective method for evaluating maps was devised that was based on the expert's heuristics. The method was successful in identifying optimal maps. A new search technique based on a form of genetic algorithm(GA) was developed to generate potential maps from a set of experimental data. The objective system for evaluating maps was incorporated into the GA. Optimal gene maps could be generated automatically, then merged together to produce a multi-gene map. In many cases, the sequence of genes and restriction enzymes was very close to the sequence as determined manually by the geneticist but could be produced in a fraction of the time.

In the automated inspection of components for surface defects using machine vision techniques, illumination must be tightly controlled in order to obtain images which highlight defective areas. This means that lighting and viewing arrangements tend to be dedicated to a particular task and are not readily adaptable to the more general case where component geometries and surface reflectivities may vary. This thesis describes the design and development of an intelligent robot vision system for the in-process monitoring of complex geometry component surfaces during abrasive polishing. Together with the use of a pair of robot manipulators as the orientating device and a periscopic lighting and viewing arrangement, the system is able to recognise and avoid those conditions which normally give rise to unsuitable images. An important feature of the system concerns the use of a computer object model of component surface geometry to assist in the inspection process. The model is derived from automated component measurements using a novel triangulation technique. The model permits an intelligent selection of viewing geometry and helps to overcome the difficult problem of interpreting images from doubly curved surfaces with non- uniform illumination. To minimize inspection cycle times an off-line optimal selection of views for the inspection of the complete object is required. This information is held in a goal oriented database which contains all the data required during in-process inspection. As the intensity variations within an image are constrained to correlate with the optical and topographical properties of the surface an initial rapid discrimination of surface defects by global grey level thresholding is possible. This is reinforced by a more sophiscated second stage analysis towards the end of the machining cycle. This employs the density of textural edges to determine the acceptability of the surface.

No description available.

No description available.

No description available.

No description available.

Ticker-2 is a Qualitative Model of cardiac electrophysiology. One of the motivations was a desire to test and improve Qualitative Modelling techniques by their application to a real-world problem. The approach used is component-based, but, unlike much other work in Qualitative Modelling, discrete valued parameters are used. The application of temporal constraints play an important part, and the Relative Endpoint Duration formalism has been developed to constrain the generation of behaviours by the application of knowledge about the domain. This formalism is unusual in Qualitative Modelling in that an explicit 'unknown' value is used, instead of allowing uncertain values to lead to nondeterministic branching. Both simulation and envisionment have been used to generate behaviours from Ticker-2. Problems exist with both, but envisionment being more appropriate to the domain of cardiac electrophysiology, a number of ways of overcoming difficulties with envisionment are developed. An approach to comparing the behaviours of Ticker-2 with KARDIO behaviours is also suggested.

No description available.

No description available.

No description available.