In this paper, we demonstrate a new approach for the fusion of multichannel signals. We show how this method can be used to combine signals from magnetometer and gradiometer sensors used in magnetoencephalography (MEG). This approach works by assuming that the lead-fields have multiplicative errors which in turn leads to an under-determined problem. To solve this problem, we impose two constraints that result in closed-from solutions: i) one set of sensors is error-free, ii) the norm of the multiplicative error is bounded. These prior assumptions to estimate the error are used in the linearly constraint minimum variance (LCMV) spatial filter to improve the optimisation. Although we focus on the fusion of MEG sensors, this approach can be employed for multimodal fusion of other multichannel signals such as MEG and EEG signals.

Evidence suggests that magnetoencephalogram (MEG) data have characteristics with non-Gaussian distribution, however, standard methods for source localisation assume Gaussian behaviour. We present a new general method for non-Gaussian source estimation of stationary signals for localising brain activity in the MEG data. By providing a Bayesian formulation for linearly constraint minimum variance (LCMV) beamformer, we extend this approach and show that how the source probability density function (pdf), which is not necessarily Gaussian, can be estimated. The proposed non-Gaussian beamformer is shown to give better spatial estimates than the LCMV beamformer, in both simulations incorporating non-Gaussian signal and in real MEG measurements.

In this study, a robust minimum variance beam-former (RMVB) is employed for source reconstruction in simultaneous MEG and local field potential (LFP) measurements. RMVB selects the electrical activity only from a specified volume of the brain while suppressing the rest. To improve imaging we added two more terms to the RMVB: the first term minimises the mean squared error (i.e. maximises the correlation) between the recorded LFP and reconstructed time courses from MEG data, the second term nulls the large inference induced by deep brain stimulation (DBS) device. A solution of this problem, relevant both with and without extra terms, is presented using the Lagrange multiplier method. This solution-if we ignore the new terms-has a simpler secular equation compared to its original solution and therefore is faster. The method is validated using both simulated and real data to show its potential use in practical applications.

Increased accuracy of sonar ranging shows deficiencies in the specular models. A new model is developed which includes diffusion reflections and is based on ray tracing methods. Results are presented which show that this model conforms more closely to experimental data.

The typical use of ultrasonic sensors has been limited to estimation of the location of targets in a robot workspace. CTFM sonars have also been used successfully in classifying primitive targets. In this paper the classification is extended to include textures typical of these found in pathways the robot may need to follow or identify. The pathway classes examined are considered to be plane surfaces of various roughness corresponding to hard smooth floor, carpet, and asphalt. Each class is modeled using an extension of the Kirchhoff approximation method describing the scattering of the acoustic wave on rough surfaces. The CTFM sonar image corresponding to each class is derived and compared with the experimental one. Then a feature is extracted that exploits the differences between the three surface models. A neural network is trained for recognition with excellent results.

Ultrasonic sensors have been widely used as time-of-flight range finding systems in mobile robots. Different variations of this scheme lead to robust identification of simple reflector types, like walls, corners and edges. In this paper an alternative approach is attempted, able not only to locate and identify simple reflectors, but to detect and recognize more complicated objects. A more sophisticated sensor, the CTFM sonar, produces an image which corresponds to a range map. The image provides information about the location and type of the reflector. A reflectivity model for planes, corners, and edges is presented and compared with some experimental results. A method to distinguish well structured reflectors from complex objects is then described and the application of the system in a room mapping task is demonstrated.

Presents three novel techniques for enhancing the power of a genetic algorithm (GA) used to design fuzzy systems: a new context-dependent coding (CDC) technique, a simple chromosome reordering operator to maximize efficiency, and the coevolution of controller set tests to force competence in all areas of state space. These measures are shown to lead to a considerable improvement over conventional GAs when used to design controllers for a standard problem, such as the cart-pole problem. We use an analysis of GAs by L. Altenberg (1994) to determine a performance measure that demonstrates that our coding scheme and reordering operator improve the ability of the GA to organize itself and evolve chromosomal structures that not only produce high scores, but improve the search efficiency of the genetic operators. We investigate the algorithm in a controller to provide parallel parking maneuvers for mobile robots. It is shown that the controllers developed are robust to the systematic errors that inevitably arise when controllers are transferred from a simulated environment to the real world.

This paper describes techniques to identify various features in a scene using a laser range finding sensor, scanning horizontally in one dimension. The authors describe the optical and electronic properties of the sensor, together with the distributed processing hardware. The authors present a method to recognise and parameterise straight lines and ellipses occurring together in ID range images. In particular, the authors: (1) Present a line fitting system which recursively fits points to line segments, given range and bearing information, and correctly accounts for the sensors error sources. This system also passes on unfitted points to the ellipse fitter. (2) Extend Rothwell and Zisserman's improvements to the Bookstein algorithm to determine ellipse parameters from few observed points which lie over a small angular excursion of the ellipse, for example a pipe in an industrial environment. (3) Discuss the problems of outliers, and develop methods of recognising and rejecting them, including the use of intensity as well as range data. The authors show solutions to these problems on real data, and finally discuss further methods they might use to improve the ellipse fitting.

A modular architecture used for intelligent control of mobile robots has been developed at Oxford University over the last few years. This architecture takes the form of multiple sensing and control layers, based on "locally intelligent control agents" (LICAs). Central to such a design is the concept of "communication sequential processes" (CSP). This paper describes briefly the LICA-based control architecture for an Oxford mobile robot. Two typical operations of sonar and active vision systems have been given to demonstrate its flexibility and efficiency. Such a modular and unified system has been licensed commercially and allows for the integration of all kinds of sensors and actuators to meet future needs.

We describe a cost-effective method for developing parallel architectures which increase the performance of range and image sensors. A parametrised edge detector and its systolic implementation using field-programmable gate arrays (FPGAs) are presented. Experiments and analyses indicate that our circuits can satisfy the performance requirements, and some of the designs out-perform the software equivalent on a 486-based PC by nearly two orders of magnitude.>

Investigates how a car-like mobile robot handles unexpected static obstacles while following an optimal path planned by the global path planner. To find an optimal solution of the problem, the obstacle avoidance problem is formulated as a decision theoretic approach. The optimal decision rule we seek is to minimize the Bayes risk by trading off between deliberative maneuver and the alternatives. Real-time implementation is emphasized in order to provide a framework for real-world applications.

Desigii considerations arc presented for rarlgefiilders wliiclii ernploy active triaiigulatinri. Tlie geometric tradeoffs for coiiveritiorial triangulation gcoirietries are identified, aiid the way in wliicli tliesc can be iiiiiiirnised using a synchronized scaiiiiiiig approach is dcscribed. We compare the perfurlrialice of the cliargecoupled device (CCD) and lateral-effect pliotocliode (LEP) as active rangefinder iriiage seiisors mid discuss the benefits arid drawbacks of different optical !sniirces (lasers arid LEDs). We predict the pcrfnrinance of a large depth of field laser/LEP seiisor arid coinpare real results with theoretical prediction.

Presents a unified approach to the navigation and control of a mobile robot. In the past, path planning has often been referred to as a 'high level' task and has been completely separated from the so called 'lower level' control of a real mobile vehicle. The authors consider here the total energy of a mobile vehicle when influenced under a good seeking navigation strategy. This energy function is used to produce a control law directly to drive a mobile vehicle. The authors also incorporate directly an estimate of an artificial repulsive potential field into the low level controller.>

Describes a decision theoretic approach to real-time obstacle avoidance and path planning for a mobile robot. The mobile robot navigates in a semi-structured environment in which unexpected obstacles may appear at random locations. Twelve sonar sensors are currently used to report the presence and location of the obstacles. To handle the uncertainty of an obstacle's appearance, the authors adopt a Bayesian approach by assuming a prior distribution for the presence of unknown obstacles. The distribution is changed dynamically according to the information accumulated by sensors. When searching for an optimal path using dynamic programming, the authors take the probability into account in making a decision. Based on prior information and sensor data, they show that the proposed method allows the mobile robot to avoid unexpected obstacles and finds an optimal path to the goal in real time.>

An architecture for the control of a multi-sensor mobile robot is proposed. The architecture is based on the subsumption approach, but with limited sharing of data between behaviours. The mechanisms were developed with the aid of the formal CSP notation, and critical parts have been proved deadlock free.>

The design and partial implementation of a real-time architecture for a mobile robot, aimed particularly towards a vehicle developed for factory automation, is described. The authors develop a layered design to equip the robot with a number of behavioral competences. They examine sensing and a potential field algorithm especially to achieve modification of behavior at a speed close to the robot's operational speed. It is shown how the layered architecture interfaces to the original onboard architecture, which provided sophisticated localization but no ability to deal with environmental exceptions.>