The imperative for global transformation towards energy sustainability leads to the increasing incorporation of photovoltaic generation into the power systems networks. The increase of photovoltaic (PV) capacity in power system, however, would reduce the system inertial response and may negatively affect the transient stability of the generators. This paper, therefore, proposes a deep reinforcement learning (DRL)-based controller to ensure the transient stability of the power grid with PV penetration. The agent of the proposed Deep Q-Network optimally learns the control actions and adjusts the excitation voltage to maintain the voltage and synchronism in the event of faults. The performance of the controller under different levels of PV penetration and LVRT capabilities are also investigated. The results are compared with that of power systems stabiliser (PSS) to verify the effectiveness. The results show that Deep Q-Network is superior to PSS in maintaining transient stability. Furthermore, the DQN is able to quickly damp out the transient oscillations at various levels of PV penetration and tripping conditions.

Real-time CPSs using Artificial Neural Networks (ANNs) are traditionally developed as monolithic black-boxes. This results in designs that are often difficult to formally verify against safety specifications and implement on hardware for formal timing analysis. Consequently, their implementation as a composition of smaller ANNs has received recent interest. These are easier to implement, parallelise and validate. Despite this, the question of how to produce hardware-implementable compositional designs from existing monolithic ones remains largely unanswered. This work develops a novel procedure to replace large ANN monolithic designs with smaller compositional designs and implement them on a Field Programmable Gate Array (FPGA) for timing analysis using synchronous compositional semantics. To illustrate our approach, we develop regression and classification ANN designs for multiple real-life datasets. Using various design and model architecture variations, we show that using a compositional design instead of a monolithic design can achieve an $\mathbf{8 5 \%}$ reduction in WCET, around a $\mathbf{53 \%}$ reduction in hardware resources and around a 40% reduction in computations and neuron connections for a minor reduction in performance.

Models have been formulated to emulate various biological cells’ action potentials (AP). Most of these are computationally expensive and unsuitable for FPGA implementations. The Resonant Model (RM) is an alternative that offers good accuracy with real-time FPGA implementation. This WIP charts the RM validation path for tissue level emulation.

Automatic detection of cardiac arrhythmia is an important tool in the fight against cardiovascular diseases and their associated human impacts. Such detection needs to be both accurate and timely, in order to allow for interventions to be administered within short time frames. Traditionally, such approaches have used black box implementations which are not explainable and hence have limited use in terms of clinical interpretability. Additionally, these implementations may either require additional training between patients, or have processing times which make them unsuitable for real-time classification. To address this, we develop a set of formal Timed Automaton-based policies that capture three common arrhythmia, Premature Ventricular Contraction, Ventricular Tachycardia, and Atrial Fibrilation, in terms of Electrocardiogram (ECG) features. We synthesise Runtime Verification monitors for each of these policies, and run them alongside existing clinical ECG databases to evaluate their efficacy. This approach shows comparable results to existing black box work with accuracies ranging from 90 % to 96 % while still being both explainable and clinically interpretable.

A well known and well-studied feature of boats’ dynamic is the effect of steering-induced roll. This property is used by a technique called Rudder Roll Stabilization (RRS) to stabilise ships in waves in order to make the navigation safer and more pleasant. This technique is based on the generation of induced roll. Because of its specific application, studies have been limited to commercial vessels using a Single Propeller-Rudder System (SPRS). This study not only broadens the technique to any propulsion and steering mechanisms that can be used with RRS by introducing the thrust asymmetry, but it also incorporates the effect of the centrifugal forces that were previously left off. To prove the capabilities of the new concept, a test is effectuated employing an RC demonstrator fitted with a Differential Jet Pump System (DJPS), performing a turning circle test manoeuvre.

This paper proposed a two-stage objective function with genetic algorithm (GA) based tuning on two controller schema - a PD/cascaded P-PD control on a traditional ball and beam system, serving as a prototype for a friction fruit conveyor system. The ball and beam system governed under such controllers is excited with step input, the corresponding system performance factors are captured - rise time, settling time and overshoot. A probabilistic random search on optimum controller parameters is carried with GA method, multiple cost functions - ISE, IAE, ITSE and ITAE, with are evaluated to form a performance cost matrix, which is the first stage of the objective function. The optimum parameter search stops with two conditions; one is that the maximum number of chromosome generation is reached, and the other one is that the performance cost stops improving consecutively for ten generations. The second stage of the objective function is proceeded to decide the found optimum controller parameter solution. The result is decided by taking account of the previously captured performance factors. These factors are normalized and combined with a heuristic weight set to determine a minimum decision cost. The minimum cost chromosome, with ITSE, is the optimum from global found solution space. With two-stage objective function, the GA tuned cascaded P-PD control result on the ball and beam system, meets all system requirements and is satisfactory. Such method can be later implemented on the friction fruit conveyor system for fruit position control and sorting applications.

This paper proposed a two-stage objective function with genetic algorithm (GA) to refine an additional fuzziness layer on dynamic membership intervals of Type-I fuzzy logic control(FLC). The refined dynamic membership intervals Type-I FLC is applied on a traditional ball and beam system, serving as a prototype for a friction fruit conveyor system. The ball and beam system governed under Type-I FLC is excited with step input, the corresponding system performance factors are captured - rise time, settling time and overshoot. A probabilistic random search on optimum controller parameters is carried with GA method, multiple cost functions - ISE, IAE, ITSE and ITAE, with are evaluated to form a performance cost matrix, which is the first stage of the objective function. The optimum parameter search stops with two conditions; one is that the maximum number of chromosome generation is reached, and the other one is that performance cost stops improving consecutively for ten generations. The second stage of the objective function is proceeded to decide the found optimum controller parameter solution. The result is decided by taking account of the previously captured performance factors. These factors are normalized and combined with a heuristic weight set to determine a minimum decision cost. The minimum cost chromosome, with ITAE, is the optimum from global found solution space and sets fixed intervals on the dynamic membership range of Type-I FLC. With two-stage objective function, improved rise time and settling time performance are indicated on the GA tuned Type-I FLC dynamic membership intervals on the ball and beam system than the conventional Type-I FLC, and is satisfactory.

OBJECTIVE: Cardiovascular Implantable Electronic Devices (CIEDs) are used extensively for treating life-threatening conditions such as bradycardia, atrioventricular block and heart failure. The complicated heterogeneous physical dynamics of patients provide distinct challenges to device development and validation. We address this problem by proposing a device testing framework within the in-silico closed-loop context of patient physiology. METHODS: We develop an automated framework to validate CIEDs in closed-loop with a high-level physiologically based computational heart model. The framework includes test generation, execution and evaluation, which automatically guides an integrated stochastic optimization algorithm for exploration of physiological conditions. CONCLUSION: The results show that using a closed loop device-heart model framework can achieve high system test coverage, while the heart model provides clinically relevant responses. The simulated findings of pacemaker mediated tachycardia risk evaluation agree well with the clinical observations. Furthermore, we illustrate how device programming parameter selection affects the treatment efficacy for specific physiological conditions. SIGNIFICANCE: This work demonstrates that incorporating model based closed-loop testing of CIEDs into their design provides important indications of safety and efficacy under constrained physiological conditions.

The present study applies a novel two-dimensional learning framework (2D-UPSO) based on particle swarms for structure selection of polynomial nonlinear auto-regressive with exogenous inputs (NARX) models. This learning approach explicitly incorporates the information about the cardinality (i.e., the number of terms) into the structure selection process. Initially, the effectiveness of the proposed approach was compared against the classical genetic algorithm (GA) based approach and it was demonstrated that the 2D-UPSO is superior. Further, since the performance of any meta-heuristic search algorithm is critically dependent on the choice of the fitness function, the efficacy of the proposed approach was investigated using two distinct information theoretic criteria such as Akaike and Bayesian information criterion. The robustness of this approach against various levels of measurement noise is also studied. Simulation results on various nonlinear systems demonstrate that the proposed algorithm could accurately determine the structure of the polynomial NARX model even under the influence of measurement noise.

The present study investigates a friction fruit conveyor system development based on a traditional friction-less ball and beam system which share the commonalities of controlling object according to platform angle. Given that the ball and beam system is inherently open-loop unstable, a simple PID controller was designed to stabilize the ball to a predefined position on the beam. In most of the ball and beam literature, the system is assumed to be ideal, friction-free and usually linearized to a simplified model. The analytical model cannot be accurate in real life application. Subsequently, system identification is a standard procedure to estimate its corresponding model for optimal controller designs. With insight from an identified state-space model, parameters such as the number of tapped delay lines and hidden layers are designed. A grey-box neural network system identification with transfer learning is then proposed to identify a nonlinear friction ball and beam system. The identified model is adaptively based on a pre-trained neural network obtained from a linear friction-free BBS mathematical system. The performance of the grey-box identified neural network system model with transfer learning, is then compared with a model obtained from its black-box identified model with neural network structure. Subsequently, a similar procedure will be used to design a grey-box neural network model for fruit conveyor system based on this friction ball and beam. The results of simulation grey-box neural network with transfer learning based on the developed friction ball and beam model system model, is satisfactory.

A new activation function is proposed. This activation function uses the square operator to introduce the required non-linearity as compared with the use of an exponential term in the popular TanSig. Smaller computational operation count characterizes the proposed activation function. The key to the effectiveness of this function is a faster convergence when used in Multilayer Perceptron Artificial Neural Network architectural problems. Besides, the derivative of the function is linear, resulting in a quicker gradient computation. The effectiveness and efficiency of the proposed activation function have been compared with the TanSig and the computationally efficient ElliotSig functions using selected UCI datasets. The MNIST handwritten dataset was also used to show the ability of this function on more massive datasets. An empirical comparison suggests that the proposed function outperforms TanSig and ElliotSig in convergence time as well as in generalization metrics for most datasets.

The heart is a vital organ that relies on the orchestrated propagation of electrical stimuli to coordinate each heartbeat. Abnormalities in the heart’s electrical behaviour can be managed with a cardiac pacemaker. Recently, the closed-loop testing of pacemakers with an emulation (real-time simulation) of the heart has been proposed. This enables developers to interrogate their pacemaker design without having to engage in costly or lengthy clinical trials. Many high-fidelity heart models have been developed, but are too computationally intensive to be simulated in real-time. Heart models, designed specifically for the closed-loop testing of pacemaker logic, are too abstract to be useful for the testing of pacemaker implementations. In the context of pacemaker testing, compared to high-fidelity heart models, this article presents a more computationally efficient heart model that generates realistic piecewise continuous electrical signals. The heart model is composed of cardiac cells that are connected by paths. Our heart model is based on the Stony Brook cardiac cell model and the UPenn path model, and improves them by stabilising the activation behaviour of the cells and by capturing the piecewise continuous behaviour of electrical propagation. We provide simulation results that show our ability to faithfully model a range of arrhythmias, such as VA conduction, heart blocks, and long Q-T syndrome. Moreover, re-entrant circuits (that cause arrhythmia) can be faithfully modelled, which only the discrete-event UPenn heart model is also able to achieve.

We describe an efficient GPU algorithm which extracts multiple contours from an image. The algorithm uses crack codes to generate contours which sit logically between adjacent image values; it works scan line by scan line and it can generate multiple contours in parallel with an image streamed directly from a camera. Whilst specifically targeted at detecting object contours in stereo disparity maps, it can also be used for general segmentation with a trivial change to the code generating the crack code masks. Using a480 ALU 1.4 GHz nVidia GPU, it can generate ~ 25000 contours from a real 2048 × 768 resolution 128 level disparity map image in ~ 29 ms if the contours are further processed in the GPU (additional ~5 ms to calculate shape moments) or ~ 39 ms if contours are transferred to the host. This is ~ 40 times faster than an OpenCV CPU implementation.

The control of a brushless dc (BLDC) motor, using bit-streams is presented. Bi-polar signals are represented using a uniformly weighted bit-stream which can be manipulated using simple digital logic to create a variety of control actions. This paper presents the functional elements required for the implementation of a proportional integral (PI) controller to affect speed and torque control on a BLDC motor. Preliminary experimental results on a standard BLDC motor using an Altera field programmable gate array (FPGA) are presented and show the viability of this technique.

A new method for the parallel hardware implementation of artificial neural networks (ANNs) using digital techniques is presented. Signals are represented using uniformly weighted single-bit streams. Techniques for generating bit streams from analog or multibit inputs are also presented. This single-bit representation offers significant advantages over multibit representations since they mitigate the fan-in and fan-out issues which are typical to distributed systems. To process these bit streams using ANNs concepts, functional elements which perform summing, scaling, and squashing have been implemented. These elements are modular and have been designed such that they can be easily interconnected. Two new architectures which act as monotonically increasing differentiable nonlinear squashing functions have also been presented. Using these functional elements, a multilayer perceptron (MLP) can be easily constructed. Two examples successfully demonstrate the use of bit streams in the implementation of ANNs. Since every functional element is individually instantiated, the implementation is genuinely parallel. The results clearly show that this bit-stream technique is viable for the hardware implementation of a variety of distributed systems and for ANNs in particular.

The present study proposes an alternate method of structure selection or which terms to include into a nonlinear autoregressive moving average with exogenous inputs (NARMAX) model, based on evolutionary programming (EP). The algorithm uses a strategy similar to elitism where the single best chromosome in a generation is retained and passed to the next. In addition to minimizing the mean square error (MSE), the method introduces an internal term penalty (ITP) function to reject spurious terms under the effects of significant noise. By following an adaptive mutation rate and restricting this to vary within 50%, faster convergence is achieved. To further improve the convergence, a pruning strategy is followed where any insignificant terms are removed from the model by assigning them with a time-to-live parameter. The performance of the proposed method is illustrated considering several examples of nonlinear systems and have been found to be satisfactory

The number of students enrolled at Auckland University School of Engineering in general and in the Department of Electrical and Electronic Engineering, in particular, has increased significantly. Different methods of learning and assessment that help both the students and the educators need to be explored. At the Department of Electrical and Electronic Engineering, OASIS (online assessment integrated system), a computer-based tool was developed to meet those needs. Its successful usage at the introductory level encouraged the authors to use it in a more advanced course. A survey was conducted to establish the usefulness of OASIS. The results of this survey show that OASIS is useful tool both for the students as well as for the educators. The students find that it assists them in the learning process by encouraging them to solve more problems. Also, the assessment in advanced and design style courses could be improved with inclusion of an OASIS component in the course.