Niall O' Mahony | Institute of Technology, Tralee (original) (raw)

Papers by Niall O' Mahony

The emergence of trends such as the Industrial Internet of Things (IIOT), Industry 4.0 and Proces... more The emergence of trends such as the Industrial Internet of Things (IIOT), Industry 4.0 and Process analytical technology has heightened the demand for real-time, in-line and affordable sensors. The mass flow rate of bulk solids is a parameter that is invaluable to process understanding in many applications. Despite this the availability of affordable sensors is limited. This paper will investigate some novel methods sensing arrangements utilising cost-effective components including optical and acoustic sensors techniques with a comparison of these methods against a commercial sensor for their evaluation. Experimental results demonstrate that reliable monitoring of powder flow parameters is achieved and that the system is able to track fluctuations of powder flow in pipes in both freefall and pneumatic conveyance applications.

Autonomous navigation is one of the most challenging competencies required of a mobile robot. In ... more Autonomous navigation is one of the most challenging competencies required of a mobile robot. In order to accomplish successful navigation, a mobile robot must be competent in the four main elements of autonomous navigation: perception- the robot must be capable of interpreting its sensors to configure useful data about its environment; localization- the robot must be capable of determining its state within that environment; cognition- the robot must be make meaningful decisions on its actions in order to achieve its goals; and motion control- the robot must be capable of modulating its motor outputs to accurately achieve its desired trajectory. Of these four elements, localization has received the most attention by researchers in recent years, and as a result, we are seeing tremendous advances being made. This paper will provide an overview of the most commonly used localization techniques for mobile robots. We highlight the advantages and challenges associated with each technique and also investigate the various sensor fusion approaches that are being applied to enhance the overall accuracy and reliability of the localization system.

arXiv (Cornell University), Aug 11, 2022

Fine-Grained Change Detection and Regression Analysis are essential in many applications of Artif... more Fine-Grained Change Detection and Regression Analysis are essential in many applications of Artificial Intelligence. In practice, this task is often challenging owing to the lack of reliable ground truth information and complexity arising from interactions between the many underlying factors affecting a system. Therefore, developing a framework which can represent the relatedness and reliability of multiple sources of information becomes critical. In this paper, we investigate how techniques in multi-task metric learning can be applied for the regression of fine-grained change in real data. The key idea is that if we incorporate the incremental change in a metric of interest between specific instances of an individual object as one of the tasks in a multi-task metric learning framework, then interpreting that dimension will allow the user to be alerted to fine-grained change invariant to what the overall metric is generalised to be. The techniques investigated are specifically tailored for handling heterogeneous data sources, i.e. the input data for each of the tasks might contain missing values, the scale and resolution of the values is not consistent across tasks and the data contains non-independent and identically distributed (non-IID) instances. We present the results of our initial experimental implementations of this idea and discuss related research in this domain which may offer direction for further research.

Springer eBooks, Jul 13, 2021

This paper presents a NIRS based real time continuous monitoring of powder blend composition whic... more This paper presents a NIRS based real time continuous monitoring of powder blend composition which has widespread applications such as the pharmaceutical industry. The paper extends the implementation of several machine learning methodologies applied to sensor data collected using an NIR spectrometer for a model powder blending process. Several techniques were examined for the development of chemometric models of the multi-sensor data, including Principal Component Analysis (PCA), Partial Least Squares Regression (PLSR), Support Vector Machines (SVM) and Artificial Neural Networks (ANN). The performances of each of the models were compared in terms of accuracy (MSE) in predicting blend composition. The results obtained show that machine learning-based approaches produce process models of similar accuracy and robustness compared to models developed by PLSR while requiring minimal pre-processing and also being more adaptable to new data. The paper also discusses the prospect of using Convolutional Neural Networks (CNN) for NIRS data analysis.

Springer eBooks, Aug 25, 2020

Deep Metric Learning (DML) approaches learn to represent inputs to a lower-dimensional latent spa... more Deep Metric Learning (DML) approaches learn to represent inputs to a lower-dimensional latent space such that the distance between representations in this space corresponds with a predefined notion of similarity. This paper investigates how the mapping element of DML may be exploited in situations where the salient features in arbitrary classification problems vary over time or due to changing underlying variables. Examples of such variable features include seasonal and time-of-day variations in outdoor scenes in place recognition tasks for autonomous navigation and age/gender variations in human/animal subjects in classification tasks for medical/ethological studies. Through the use of visualisation tools for observing the distribution of DML representations per each query variable for which prior information is available, the influence of each variable on the classification task may be better understood. Based on these relationships, prior information on these salient background variables may be exploited at the inference stage of the DML approach by using a clustering algorithm to improve classification performance. This research proposes such a methodology establishing the saliency of query background variables and formulating clustering algorithms for better separating latent-space representations at run-time. The paper also discusses online management strategies to preserve the quality and diversity of data and the representation of each class in the gallery of embeddings in the DML approach. We also discuss latent works towards understanding the relevance of underlying/multiple variables with DML.

Advances in intelligent systems and computing, Nov 8, 2018

This paper will review the progress which has been made in Artificial Intelligence and Computer V... more This paper will review the progress which has been made in Artificial Intelligence and Computer Vision particularly in 3D computer vision. There has been a lot of activity in the development of both hardware and software in 3D imaging systems which will have a huge impact in the capabilities of robotics. This paper reviews the latest advancements in the state of the art in range imaging sensors as well as some emerging technologies. For example, Time of Flight (ToF) cameras with improved resolution and latency, low cost LiDAR, and the fusion of range imaging technologies will empower robotics with greater perception capabilities. Likewise, software approaches will be reviewed with a focus on Deep Learning approaches which are now the leading edge in data analysis and further enhancing the capabilities of intelligent robotic systems using 3D imaging. The emergence of Geometric Deep Learning for 3D computer vision in robotics will also be detailed, with a focus on object registration, object detection and semantic segmentation. Foreseeable trends which have been identified in both hardware and software aspects of 3D computer vision are also discussed.

Procedia Computer Science, 2019

Abstract This research investigates the impact of edge agents on industrial plants in the era of ... more Abstract This research investigates the impact of edge agents on industrial plants in the era of the Internet of Things (IoT) and the increasing availability of internet connection. This paper proposes ‘Edge Agent’ a holistic solution to managing many devices on the edge and will give a brief introduction to the communication between agents and existing machinery as well as present results which were extracted from experiments performed with our solution under low load in terms of data and with a small number of devices in terms of distribution. As result of extensive architecture investigation for an optimal edge solution and its possible correlation to industrial applications, this paper will introduce edge agents, communication between agents and machinery and industrial applications. The paper will present some important findings on edge computing, compare main architectural aspects and will provide a broad view of how edge solutions might be built for this particular scenario. Having discussed how the ideal architecture works and having provided an overview about how it may be applied to industrial plants, the final section of this paper addresses how artificial intelligence will fit into edge solutions, introducing important trends like the one-shot-learning technique, forming a new source of “smart capabilities” to existing environments.

Advances in intelligent systems and computing, 2020

Deep Learning has pushed the limits of what was possible in the domain of Digital Image Processin... more Deep Learning has pushed the limits of what was possible in the domain of Digital Image Processing. However, that is not to say that the traditional computer vision techniques which had been undergoing progressive development in years prior to the rise of DL have become obsolete. This paper will analyse the benefits and drawbacks of each approach. The aim of this paper is to promote a discussion on whether knowledge of classical computer vision techniques should be maintained. The paper will also explore how the two sides of computer vision can be combined. Several recent hybrid methodologies are reviewed which have demonstrated the ability to improve computer vision performance and to tackle problems not suited to Deep Learning. For example, combining traditional computer vision techniques with Deep Learning has been popular in emerging domains such as Panoramic Vision and 3D vision for which Deep Learning models have not yet been fully optimised.

Mobile robots have become increasingly popular in recent years, offering a wide range of applicat... more Mobile robots have become increasingly popular in recent years, offering a wide range of applications in areas such as industry, agriculture, search and rescue and much more. This has been achieved mainly as a result of extremely active research and development work on robotic and autonomous technology. We are still faced with many challenges however in order for a robot to navigate efficiently and reliably in an environment without any human assistance. The robot should be capable of extracting the necessary information from the environment and taking the necessary action required to plan a feasible path for collision free motion to reach its goal. In this paper, we review the most commonly used path planning methodologies that have been applied for mobile robot navigation in both static and dynamic environments. We look at both global and local path planning approaches as well as classical and heuristic based techniques.

International journal of sustainable energy development, Dec 1, 2016

Increased globalisation and competition are drivers for process analytical technologies (PAT) tha... more Increased globalisation and competition are drivers for process analytical technologies (PAT) that enable seamless process control, greater flexibility and cost efficiency in the process industries. The article will discuss process modelling and control for industrial applications with an emphasis on solutions enabling the real-time data analytics of sensor measurements that PAT demands. This research aims to introduce an integrated process control approach, embedding novel sensors for monitoring in real time the critical control parameters of key processes in the minerals, ceramics, non-ferrous metals, and chemical process industries. The article presents a comparison of machine learning algorithms applied to sensor data collected for a polymerisation process. Several machine learning algorithms including Adaptive Neuro-Fuzzy Inference Systems, Neural Networks and Genetic Algorithms were implemented using MATLAB® Software and compared in terms of accuracy (MSE) and robustness in modelling process progression. The results obtained show that machine learningbased approaches produce significantly more accurate and robust process models compared to models developed manually while also being more adaptable to new data. The article presents perspectives on the potential benefits of machine learning algorithms with a view to their future in the industrial process industry.

The domain of 3D Deep learning is growing rapidly as 3D sensor cost plunges and the perception ca... more The domain of 3D Deep learning is growing rapidly as 3D sensor cost plunges and the perception capabilities these sensors can provide is continuously being extended. Dataset creation and annotation is a huge bottleneck in this field of work however, particularly in 3D segmentation tasks where every point in 3D space must be labelled accurately. This paper will review some creative ways of improving the data annotation process in terms of efficiency, accuracy and automatability. The review is comprised of two halves, firstly, annotation tools which have improved the user interface for pointcloud annotation are presented including works which use technologies such as virtual reality. Secondly, automation schemes which delegate as much of the work as possible to a machine while still giving the user insight and control over the process will be reviewed.

Increased globalisation and competition are drivers for process analytical technologies (PAT) tha... more Increased globalisation and competition are drivers for process analytical technologies (PAT) that enable seamless process control, greater flexibility and cost efficiency in the process industries. The paper will discuss process modelling and control for industrial applications with an emphasis on solutions enabling the real-time data analytics of sensor measurements that PAT demands. This research aims to introduce an integrated process control approach, embedding novel sensors for monitoring in real time the critical control parameters of key processes in the minerals, ceramics, non-ferrous metals, and chemical process industries. The paper presents a comparison of machine learning algorithms applied to sensor data collected for a polymerisation process. Several machine learning algorithms including Adaptive Neuro-Fuzzy Inference Systems, Neural Networks and Genetic Algorithms were implemented using MATLAB® Software and compared in terms of accuracy (MSE) and robustness in modelling process progression. The results obtained show that machine learning-based approaches produce significantly more accurate and robust process models compared to models developed manually while also being more adaptable to new data. The paper presents perspectives on the potential benefits of machine learning algorithms with a view to their future in the industrial process industry.

This research investigates how advances in Internet of Things (IoT) and availability of internet ... more This research investigates how advances in Internet of Things (IoT) and availability of internet connection would enable Edge Solutions to promote smart utilization of existing machines at the edge. The presented results are based on experiments performed in real scenarios using the proposed solution. Whereas scenarios were cloned from real environments it is important to have in mind that experiments were performed with low load in terms of data and small number of devices in terms of distribution. As result of extensive architecture investigation for an optimal edge solution and its possible correlation to industrial applications, this paper will provide evidences supporting the use of edge solutions in challenging conditions which arise at the edge, including smart factories and smart agriculture. The present work assumes that the reader has some exposition to Edge computing, Cloud computing and software development. The paper will present some important findings on this area, compare main architectural aspects and will provide a broad view of how edge solutions might be built for this particular scenario. Having discussed how the ideal architecture works and having provided an overview about how it may be applied to industrial plants, the final section of this paper addresses how artificial intelligence will fit into edge solutions, forming a new source of “smart capabilities” to existing environments.

Czech Journal of Animal Science, Apr 30, 2020

The overall objective of this study was to improve the reproductive efficiency of lactating dairy... more The overall objective of this study was to improve the reproductive efficiency of lactating dairy cows and to improve the resulting total farm profit. The hypothesis is that a dairy farm can substantially improve its economic and environmental performance through increasing pregnancy rate, i.e. increasing the number of eligible cows that become pregnant for a given breeding period. This paper presents a tool which was designed with a view to comparing the reproductive efficiency. The tool was developed using dynamic programming in R (Shiny) and shows the changes in costs, revenues and net return projected for a given change in pregnancy rate. The model calculates from the first day in milk and stops when the last calf was born after successful insemination of each cow. Sensitivity analyses demonstrated that the economic return associated with reproductive performance is greatly affected by the input parameters and therefore real farm and market values are crucial. The average economic gain per percentage point of 21-d (21-day) pregnancy rate (PR) was 14.6 EUR per cow/year. The milk price showed the largest impact on the overall net return. A 10% increase in milk price increased the net return on average by 268 EUR (10% 21-d PR), 292 EUR (20% 21-d PR) and 299 EUR per cow/year (30% 21-d PR). Our study had the same set values of milk yield during lactations for all four evaluated farms and it was found that the milk income over feed cost increased with the reproductive performance in all evaluated farms on an individual cow level. Poor fertility means that cows spend longer producing lower amounts of less efficiently produced milk.

Procedia Manufacturing, 2019

Deep Learning has great achievements in computer vision for various classification and regression... more Deep Learning has great achievements in computer vision for various classification and regression tasks. The automation of tasks such as component sorting, bin-picking and anomaly detection may be of great use in the process industry. However, most machine learning-based object categorization algorithms require training on hundreds or thousands of images and very large datasets. The requirement for large training datasets presents a barrier to the adoption of deep learning methodologies in many custom object classification tasks. For example, in defect detection, positive instances of a defect, take for instance a tank leakage, may seldom occur and therefore creating a dataset of sufficient size for conventional deep learning procedures is not always possible. One-shot learning aims to learn information about object categories from only a handful of labelled examples per category. One-shot learning has received the most attention in face-recognition and person re-identification (re-id) tasks due to their potential practical applications in surveillance security. This research will review these one-shot learning methodologies and investigate how they may be transferred to other domains. Concepts such as Siamese Networks and triplet loss which are commonly used for one-shot learning will be examined. Challenges such as variations in illumination conditions, object pose, camera resolution and partial occlusion will be discussed. Finally, the implications and advantages of deploying such techniques to practical applications in the process industry will be analysed.

This paper will review the main sensor technologies used to create an autonomous vehicle. Sensors... more This paper will review the main sensor technologies used to create an autonomous vehicle. Sensors are key components for all types of autonomous vehicles because they can provide the data required to perceive the surrounding environment and therefore aid the decision-making process. This paper explains how each of these sensors work, their advantages and disadvantages and how sensor fusion techniques can be utilised to create a more optimum and efficient system for autonomous vehicles.

Due to the increasing availability of internet connection both industrial and agricultural enviro... more Due to the increasing availability of internet connection both industrial and agricultural environments have experienced an expressive increase in the number of IoT devices for the last few years. This paper exposes an extensive architecture investigation looking for an optimal edge solution to cope with existing, and future applications. This paper presents an introduction to the proposed solution in terms of architecture, including local user interface, MQTT broker, machine learning edge engine, some existing management software, local databases, actual software used for hardware communication, the edge agent and aspects of the hardware used during tests. It also presents some results obtained from experiments under heavy load in terms of data and a small number of devices in terms of distribution. Together the paper brings some important findings on edge computing, compare main architectural aspects and will provide a broad view of how edge solutions might be built for this particular scenario. Having discussed how the ideal architecture works and having provided an overview of how it may be applied to industrial and agricultural environments, the conclusion points to the next steps for the current research.

Increased globalisation and competition are drivers for process analytical technologies (PAT) tha... more Increased globalisation and competition are drivers for process analytical technologies (PAT) that enable seamless process control, greater flexibility and cost efficiency in the process industries. This research is carried out in collaboration with a project which aims to introduce an integrated process control approach, embedding novel sensors for monitoring in real time the critical control parameters of key processes in the minerals, ceramics, non-ferrous metals, and chemical process industries. The paper will review the development of a suite of affordable sensors along with smart sensor features and algorithms for easier integration, easier maintenance, metrological performance enhancement, process monitoring and control and sensor fusion for use within this versatile global control platform implementing PAT. Smart sensors will be investigated that match existing offline solutions in performance while enabling size reductions, low power consumption, low unit costs, low maintenance costs and data fusion.

arXiv (Cornell University), Aug 11, 2022

Springer eBooks, Jul 13, 2021

Springer eBooks, Aug 25, 2020

Advances in intelligent systems and computing, Nov 8, 2018

Procedia Computer Science, 2019

Advances in intelligent systems and computing, 2020

International journal of sustainable energy development, Dec 1, 2016

Increased globalisation and competition are drivers for process analytical technologies (PAT) tha... more Increased globalisation and competition are drivers for process analytical technologies (PAT) that enable seamless process control, greater flexibility and cost efficiency in the process industries. The paper will discuss process modelling and control for industrial applications with an emphasis on solutions enabling the real-time data analytics of sensor measurements that PAT demands. This research aims to introduce an integrated process control approach, embedding novel sensors for monitoring in real time the critical control parameters of key processes in the minerals, ceramics, non-ferrous metals, and chemical process industries. The paper presents a comparison of machine learning algorithms applied to sensor data collected for a polymerisation process. Several machine learning algorithms including Adaptive Neuro-Fuzzy Inference Systems, Neural Networks and Genetic Algorithms were implemented using MATLAB® Software and compared in terms of accuracy (MSE) and robustness in modelling process progression. The results obtained show that machine learning-based approaches produce significantly more accurate and robust process models compared to models developed manually while also being more adaptable to new data. The paper presents perspectives on the potential benefits of machine learning algorithms with a view to their future in the industrial process industry.

Czech Journal of Animal Science, Apr 30, 2020

Procedia Manufacturing, 2019

Increased globalisation and competition are drivers for process analytical technologies (PAT) tha... more Increased globalisation and competition are drivers for process analytical technologies (PAT) that enable seamless process control, greater flexibility and cost efficiency in the process industries. This research is carried out in collaboration with a project which aims to introduce an integrated process control approach, embedding novel sensors for monitoring in real time the critical control parameters of key processes in the minerals, ceramics, non-ferrous metals, and chemical process industries. The paper will review the development of a suite of affordable sensors along with smart sensor features and algorithms for easier integration, easier maintenance, metrological performance enhancement, process monitoring and control and sensor fusion for use within this versatile global control platform implementing PAT. Smart sensors will be investigated that match existing offline solutions in performance while enabling size reductions, low power consumption, low unit costs, low maintenance costs and data fusion.