We report on the application of an evolutionary algorithm to a noisy, dynamic optimization problem in chemistry: the maximization of three-photon absorption in molecular iodine. An evolution strategy is used in real-time in a closed loop experiment to search the space of physically realizable phase-modulated femtosecond laser pulses. The probability of three-photon absorption is estimated by measuring UV fluorescence. With the evolutionary search it is possible to enhance the UV fluorescence by a factor of 3.4 compared to the most intense pulse.

This paper concerns the use of genetic programming (GP) for supervised classification of transcriptome (gene expression) data. In such applications GP can produce accurate predictive models that generalize well and use only very few gene expression values. It is often suggested that the selected genes are therefore of biological significance in discriminating the classes. The paper presents a preliminary study of successful parsimonious GP models to investigate the extent to which the selected variables contribute to the classification. The work is based on a readily available and well studied dataset that represents gene expression values for two groups of patients with different forms of Leukemia.

This paper relates to supervised interpretation of the infrared analytical spectra of complex biological samples. The aim is to produce a model that can predict the value of a measurand of interest, such as the concentration of a particular chemical constituent in complex biological material. Conventionally, a number of spectra are co-added to reduce measurement noise and this is time consuming. In this paper we demonstrate the ability of evolutionary search to provide adaptive averaging of spectral regions to provide selective tradeoff between spectral resolution and signal-to-noise ratio. The resultant denoised subset of the variables is then input to a proprietary Genetic Programming (GP) package which forms a predictive model that compares well in predictive power with a combination of Partial Least Squares Regression (PLS) and adaptive denoising. This demonstrates the considerable advantage that, given appropriate node functions, the GP could handle the entire process of denoising and forming the final predictive model all in one stage. This reduces or removes the need for co-adding with a consequent reduction in data acquisition time.

We present a method in which a genetic algorithm is used to optimise an expression in order to provide a supervised method for interpretation of the infrared analytical spectra of complex biological samples. The aim is to produce a model that can predict the value of a measurand of interest, such as the concentration of a particular chemical constituent, from a complex infrared spectrum of biological material. The method we describe is in some ways analogous to genetic programming but it more readily allows the output expression to be constrained in complexity and permits its general form to be specified by the user, thereby enhancing its explanatory ability. The quasi-continuous properties of optical spectra are exploited by mutations that explore spectral regions adjacent to selected variables, and provide adaptive averaging of spectral regions so as to provide selective optimisation of the tradeoff between spectral resolution and signal-to-noise ratio.

We report the results of a programme of work that allows us to specify and execute robotic assembly of snack food products such as sandwiches and pizzas simply by presenting to the system hand-prepared examples of the desired results of the robot operations. We summarise a detailed experimental evaluation of our pick and place work, in which placement location and orientation are taught by example. We also describe and evaluate our recent work on automated spreading of ingredients so as to form specific surface profiles. Our methods facilitate rapid reconfiguration to accommodate new product variants without the need for reprogramming.

Sectors of the food processing industry have challenging requirements for automation: short batches, product innovation, variation in component shape, all with high production volume. Hard-programmed automation cannot provide the flexibility required. We report our initial work on an approach in which robot operations are derived from the sensed difference between a part-assembled product and a pre-sensed example of the result required. We describe a workcell with laser sensor and present results for experiments on simple assemblies of real sandwich and pizza components. The technique offers flexibility without reprogramming.

The aim of this work is to investigate behavioural control strategies inspired by 'perceptual control theory' demonstrated by Powers (1973), and their potential application in an industrial robotic handling task. In the particular approach reported here, an Adept robot is stationed over a conveyor belt which delivers the frozen food product pieces at random position and orientation. A vision system, placed upstream of the robot, senses the position and shape of individual product items. The robot grasps each product item and transports it to a packing location where it is placed into a box. Achievement of reliable grasping is made more difficult by shape variations within individual product types, by the need to accommodate, without software modifications, new shapes as new products are developed, and by the need to change from one product type to another with minimal operator involvement.

We have developed a means of interfacing simulation systems to a manufacturing plant in such a way as to permit the use of simulation to provide real time visualization of the operation of the plant, via 3D animation. The system also permits collection of statistics relating to the manufacturing operation. Because the simulation is continually updated to reflect the current state of the plant, the simulation may also be used for predictive purposes to examine the effect of changes imposed on the operation. We have demonstrated successfully the real time monitoring technique by animating, on the basis of emulated factory floor devices, a simulation model of a sheep processing line involving multiple robots, machines, conveyors, and manual labour units. The client/server approach used is capable of supporting distributed monitoring and control across a factory network. The technique is independent of proprietary simulation packages, and so provides benefits of versatility and application independence.

This paper describes a control method for searching suitable gripping points on the outline of generic shapes. It is aimed at robotic grasping tasks where no full geometrical knowledge of the shape is assumed. We first describe the method for a range of 2D generic shapes and then instantiate the method for pick and place operations on known shapes. The algorithms are run on the shape as it appears on the computer screen directly from a vision system. Virtual robotic fingers and associated sensors are then configured and positioned on the screen with three variables being simultaneously controlled. When the control systems reach a steady state, information about virtual finger position and orientation are then used to drive the manipulator.

This paper presents the initial development of a visually guided robotic system for handling food products that are often presented in an unstructured manner. In order to handle the communications between the host and robot controller more effectively, we developed multi-level commands within the host programming environment. Vision guided grasping is described in the context of vector manipulations. The system is able to calculate the grasp vector according to both online visual information and off-line data based on the food product and gripper type in consideration. A case study is presented where the system handles trapezoidal shaped fish products.>

Advanced robotic applications often require a robot to work as a part of an integrated system. Supervisory control of the robot in such cases readily permits incorporation of external sensing systems such as vision. To handle communication between a host and robot controller more effectively, we have developed multilevel commands within the host programming environment. An example application is presented which is a visually guided robot food handling task. The work cell comprises a robot manipulator, a vision system, and a host controller whereby the supervisory control of the robot is achieved.>