Skip to main content

Hongmei He

University of Salford, School of Science, Engineering and Environments, Faculty Member

Followers

21

Following

4

Co-authors

3

Public Views

I am a Professor of AI and Robotics conducting cutting-edge research in human-centred AI for trustworthy robotics and autonomous systems (RAS). My work focuses on high-impact AI applications in Digital Healthcare and Digital Infrastructure. I bridge AI theory and real-world application by integrating emerging technologies to build systems that are safe, explainable, and usable in complex, dynamic environments.
Driven by a commitment to societal benefit, my research is inherently interdisciplinary, developing AI solutions that are transparent, inclusive, and aligned with human values. I address global challenges through the design of intelligent and ethically aligned RAS that enhance health, wellbeing, and quality of life.
My work supports the responsible adoption of AI in line with the UN Sustainable Development Goals—particularly SDG 3: Good Health and Wellbeing and SDG 11: Sustainable Cities and Communities—advancing safer healthcare, resilient infrastructure, and more equitable societies.
Address: Manchester, United Kingdom

less

InterestsView All (8)

Uploads

Papers by Hongmei He

Information gain directed genetic algorithm wrapper feature selection for credit rating

Applied Soft Computing, Aug 1, 2018

Various statistics-based machine learning techniques have been employed for this task. "Curse of ... more Various statistics-based machine learning techniques have been employed for this task. "Curse of Dimensionality" is still a significant challenge in machine learning techniques. Some research has been carried out on Feature Selection (FS) using genetic algorithm as wrapper to improve the performance of credit scoring models. However, the challenge lies in finding an overall best method in credit scoring problems and improving the time-consuming process of feature selection. In this study, the credit scoring problem is investigated through feature selection to improve classification performance. This work proposes a novel approach to feature selection in credit scoring applications, called as Information Gain Directed Feature Selection algorithm (IGDFS), which performs the ranking of features based on information gain, propagates the top m features through the GA wrapper (GAW) algorithm using

Intelligent Prediction of Dynamic Yield Stress in 3D Printing Concrete Based on Machine Learning

Advances in Engineering Technology Research, Jul 19, 2023

Applying 3D printing technology to the construction industry can bring many benefits. However, du... more Applying 3D printing technology to the construction industry can bring many benefits. However, due to the specificity of 3D printing technology, its application in the construction industry has not yet been promoted. Machine learning (ML) techniques, which are popular at this stage, are expected to provide solutions to these problems. Rheological properties have been a key parameter for the quality of 3D printing concrete, and its accurate prediction can help to integrate 3D printing technology into the construction industry. In this study, a GA-RF model for predicting the dynamic yield stress (DYS) of 3D printing concrete is proposed for the first time, and the hyperparameters of the RF model are intelligently tuned during the training process. In addition, the importance analysis of the input parameters is performed. The results show that the developed prediction model has a high accuracy and the SF content has the most significant effect on DYS. The research results help to advance the construction industry to mass production of 3D printing concrete.

Modified ultrasound scalpel haemorrhoidectomy versus conventional haemorrhoidectomy for mixed haemorrhoids: A study protocol for a single-blind randomized controlled trial

Research Square (Research Square), May 11, 2022

Background: Haemorrhoids are common and frequently occurring diseases in the clinical setting, an... more Background: Haemorrhoids are common and frequently occurring diseases in the clinical setting, and severe haemorrhoids require surgical treatment. There are various surgical methods to treat haemorrhoids, but each has its advantages and disadvantages. In recent years, ultrasonic scalpels have been used in haemorrhoid surgery and have achieved good results. Ultrasonic scalpel

A Heuristically Self-Organised Linguistic Attribute Deep Learning in Edge Computing For IoT Intelligence

arXiv (Cornell University), Jun 8, 2020

With the development of Internet of Things (IoT), IoT intelligence becomes emerging technology. "... more With the development of Internet of Things (IoT), IoT intelligence becomes emerging technology. "Curse of Dimensionality" is the barrier of data fusion in edge devices for the success of IoT intelligence. Deep learning has attracted great attention recently, due to the successful applications in several areas, such as image processing and natural language processing. However, the success of deep learning benefits from GPU computing. A Linguistic Attribute Hierarchy (LAH), embedded with Linguistic Decision Trees (LDTs) can represent a new attribute deep learning. In contrast to the conventional deep learning, an LAH could overcome the shortcoming of missing interpretation by providing transparent information propagation through the rules, produced by LDTs in the LAH. Similar to the conventional deep learning, the computing complexity of optimising LAHs blocks the applications of LAHs. In this paper, we propose a heuristic approach to constructing an LAH, embedded with LDTs for decision making or classification by utilising the distance correlations between attributes and between attributes and the goal variable. The set of attributes is divided to some attribute clusters, and then they are heuristically organised to form a linguistic attribute hierarchy. The proposed approach was validated with some benchmark decision making or classification problems from the UCI machine learning repository. The experimental results show that the proposed self-organisation algorithm can construct an effective and efficient linguistic attribute hierarchy. Such a self-organised linguistic attribute hierarchy embedded with LDTs can not only efficiently tackle 'curse of dimensionality' in a single LDT for data fusion with massive attributes, but also achieve better or comparable performance on decision making or classification, compared to the single LDT for the problem to be solved. The self-organisation algorithm is much efficient than the Genetic Algorithm in Wrapper for the optimisation of LAHs. This makes it feasible to embed the self-organisation algorithm in edge devices for IoT intelligence.

A Novel Navigation System for an Autonomous Mobile Robot in an Uncertain Environment

arXiv (Cornell University), Jun 8, 2020

In this paper, we developed a new navigation system, which detects obstacles in a sliding window ... more In this paper, we developed a new navigation system, which detects obstacles in a sliding window with an adaptive threshold clustering algorithm, classifies the detected obstacles with a decision tree, heuristically predicts potential collision and finds optimal path with a simplified Mophin algorithm. This system has the merits of optimal free-collision path, small memory size and less computing complexity, compared with the state of the arts in robot navigation. The experiments on simulation and a robot for eight scenarios demonstrate that the robot can effectively and efficiently avoid potential collisions with any static or dynamic obstacles in its surrounding environment.

RobotGPT: From ChatGPT to Robot Intelligence

ChatGPT, a language model developed by OpenAI, has triggered a new wave of AI development. This a... more ChatGPT, a language model developed by OpenAI, has triggered a new wave of AI development. This article reviews the principles of ChatGPT, analyzes five types of robot perception related to human perception (visual, auditory, tactile, olfactory, and gustatory), and seven types of robot intelligence (logical-mathematical, linguistic, spatial, musical, bodily-kinesthetic, interpersonal, and naturalistic), but excludes intrapersonal intelligence as a type of robot intelligence. Human intrapersonal intelligence is related to human self-awareness, personality, biases, and ethics, all of which should be carefully considered in robotic systems. The author proposes a multimodal approach that takes advantage of GPT-3 to implement the seven types of robot intelligence and provides a framework for robot GPT. No data is applied in this paper, except the references.

The Challenges and Opportunities of Artificial Intelligence for Trustworthy Robots and Autonomous Systems

2020 3rd International Conference on Intelligent Robotic and Control Engineering (IRCE)

The challenges and opportunities of artificial intelligence for trustworthy The challenges and op... more

Analytical Review of Cybersecurity for Embedded Systems

IEEE Access

To identify the key factors and create the landscape of cybersecurity for embedded systems (CSES)... more To identify the key factors and create the landscape of cybersecurity for embedded systems (CSES), an analytical review of the existing research on CSES has been conducted. The common properties of embedded systems, such as mobility, small size, low cost, independence, and limited power consumption when compared to traditional computer systems, have caused many challenges in CSES. The conflict between cybersecurity requirements and the computing capabilities of embedded systems makes it critical to implement sophisticated security countermeasures against cyber-attacks in an embedded system with limited resources, without draining those resources. In this study, twelve factors influencing CSES have been identified: (1) the components; (2) the characteristics; (3) the implementation; (4) the technical domain; (5) the security requirements; (6) the security problems; (7) the connectivity protocols; (8) the attack surfaces; (9) the impact of the cyber-attacks; (10) the security challenges of the ESs; (11) the security solutions; and (12) the players (manufacturers, legislators, operators, and users). A Multiple Layers Feedback Framework of Embedded System Cybersecurity (MuLFESC) with nine layers of protection is proposed, with new metrics of risk assessment. This will enable cybersecurity practitioners to conduct an assessment of their systems with regard to twelve identified cybersecurity aspects. In MuLFESC, the feedback from the system-components layer to the system-operations layer could help implement ''Security by Design'' in the design stage at the bottom layer. The study provides a clear landscape of CSES and, therefore, could help to find better comprehensive solutions for CSES. INDEX TERMS Characteristics of embedded system, countermeasures, embedded system, cybersecurity of embedded system, MuLFESC, risk assessment.

A Knowledge-Based Cognitive Architecture Supported by Machine Learning Algorithms for Interpretable Monitoring of Large-Scale Satellite Networks

Sensors

Cyber–physical systems such as satellite telecommunications networks generate vast amounts of dat... more Cyber–physical systems such as satellite telecommunications networks generate vast amounts of data and currently, very crude data processing is used to extract salient information. Only a small subset of data is used reactively by operators for troubleshooting and finding problems. Sometimes, problematic events in the network may go undetected for weeks before they are reported. This becomes even more challenging as the size of the network grows due to the continuous proliferation of Internet of Things type devices. To overcome these challenges, this research proposes a knowledge-based cognitive architecture supported by machine learning algorithms for monitoring satellite network traffic. The architecture is capable of supporting and augmenting infrastructure engineers in finding and understanding the causes of faults in network through the fusion of the results of machine learning models and rules derived from human domain experience. The system is characterised by (1) the flexibi...

A Comprehensive Obstacle Avoidance System of Mobile Robots Using an Adaptive Threshold Clustering and the Morphin Algorithm

Advances in Intelligent Systems and Computing, 2018

To solve the problem of obstacle avoidance for a mobile robot in unknown environment, a comprehen... more To solve the problem of obstacle avoidance for a mobile robot in unknown environment, a comprehensive obstacle avoidance system (called ATCM system) is developed. It integrates obstacle detection, obstacle classification, collision prediction and obstacle avoidance. Especially, an Adaptive-Threshold Clustering algorithm is developed to detect obstacles, and the Morphin algorithm is applied for path planning when the robot predicts a collision ahead. A dynamic circular window is set to continuously scan the surrounding environment of the robot during the task period. The simulation results show that the obstacle avoidance system enables robot to avoid any static and dynamic obstacles effectively.

A novel navigation system for an autonomous mobile robot in an uncertain environment

Robotica, 2021

In this paper, we developed a new navigation system, called ATCM, which detects obstacles in a sl... more In this paper, we developed a new navigation system, called ATCM, which detects obstacles in a sliding window with an adaptive threshold clustering algorithm, classifies the detected obstacles with a decision tree, heuristically predicts potential collision and finds optimal path with a simplified Morphin algorithm. This system has the merits of optimal free-collision path, small memory size and less computing complexity, compared with the state of the arts in robot navigation. The modular design of 6-steps navigation provides a holistic methodology to implement and verify the performance of a robot’s navigation system. The experiments on simulation and a physical robot for the eight scenarios demonstrate that the robot can effectively and efficiently avoid potential collisions with any static or dynamic obstacles in its surrounding environment. Compared with the particle swarm optimisation, the dynamic window approach and the traditional Morphin algorithm for the autonomous navigat...

A Novel Approach for Detecting Cyberattacks in Embedded Systems Based on Anomalous Patterns of Resource Utilization-Part I

IEEE Access, 2021

This paper presents a novel security approach called Anomalous Resource Consumption Detection (AR... more This paper presents a novel security approach called Anomalous Resource Consumption Detection (ARCD), which acts as an additional layer of protection to detect cyberattacks in embedded systems (ESs). The ARCD approach is based on the differentiation between the predefined standard resource consumption pattern and the anomalous consumption pattern of system resource utilization. The effectiveness of the proposed approach is tested in a rigorous manner by simulating four types of cyberattacks: a denial-of-service attack, a brute-force attack, a remote code execution attack, and a man-in-the-middle attack, which are executed on a Smart PiCar (used as the testbed). A septenary tuple model consisting of seven parameters, representing the embedded system's architecture, has been created as the core of the detection mechanism. The approach's efficiency and effectiveness has been validated in terms of range and pattern by analyzing the collected data statistically in terms of mean, median, mode, standard deviation, range, minimum, and maximum values. The results demonstrated the potential for defining a standard pattern of resource utilization and performance of the embedded system due to a significant similarity of the parameters' values at normal states. In contrast, the attacked cases showed a definite, observable, and detectable impact on resource consumption and performance of the embedded system, causing an anomalous pattern. Thus, by merging these two findings, the ARCD approach has been developed. ARCD facilitates building secure operating systems in line with the ES's capabilities. Furthermore, the ARCD approach can work along with existing countermeasures to augment the security of the operating system layer. INDEX TERMS Anomalous resource consumption, brute-force attack, cyberattacks, denial-of-service attack, embedded systems, password attack, remote code execution, testbed. In light of the enormous expansion and spread of embedded systems (ESs) applications in different application domains such as communication, transportation, education, defense, energy, and medicine, ESs form critical components in cyberphysical systems (CPS) and IoT-enabled systems. Industry 4.0, the industrial internet of things (IIoT), integrates all systems into the Internet, such as supervisory control and data acquisition systems (SCADA), industrial automation and The associate editor coordinating the review of this manuscript and approving it for publication was Xiali Hei . control systems (IACS), industrial control systems (ICS), and distributed control systems (DCS). Hence, cybersecurity by design is a critical challenge for the success of these modern applications. This highlights the urgent need to secure these systems against cyberattacks, which can cause catastrophic consequences to both humans and the economy. There are many mature techniques for protecting IT infrastructure. However, compared to conventional multitasking systems, ESs face significant challenges to adopt existing advanced security solutions due to their capabilities and limiting characteristics, such as small size, small volume of memory, less computational capabilities, low energy consumption,

Vulnerability-Based Impact Criticality Estimation for Industrial Control Systems

2020 International Conference on Cyber Security and Protection of Digital Services (Cyber Security), 2020

Cyber threats directly affect the critical reliability and availability of modern Industry Contro... more Cyber threats directly affect the critical reliability and availability of modern Industry Control Systems (ICS) in respects of operations and processes. Where there are a variety of vulnerabilities and cyber threats, it is necessary to effectively evaluate cyber security risks, and control uncertainties of cyber environments, and quantitative evaluation can be helpful. To effectively and timely control the spread and impact produced by attacks on ICS networks, a probabilistic Multi-Attribute Vulnerability Criticality Analysis (MAVCA) model for impact estimation and prioritised remediation is presented. This offer a new approach for combining three major attributes: vulnerability severities influenced by environmental factors, the attack probabilities relative to the vulnerabilities, and functional dependencies attributed to vulnerability host components. A miniature ICS testbed evaluation illustrates the usability of the model for determining the weakest link and setting security priority in the ICS. This work can help create speedy and proactive security response. The metrics derived in this work can serve as sub-metrics inputs to a larger quantitative security metrics taxonomy; and can be integrated into the security risk assessment scheme of a larger distributed system.

The Challenges and Opportunities of Human-Centered AI for Trustworthy Robots and Autonomous Systems

IEEE Transactions on Cognitive and Developmental Systems, 2021

The trustworthiness of Robots and Autonomous Systems (RAS) has gained a prominent position on man... more The trustworthiness of Robots and Autonomous Systems (RAS) has gained a prominent position on many research agendas towards fully autonomous systems. This research systematically explores, for the first time, the key facets of humancentered AI (HAI) for trustworthy RAS. In this article, five key properties of a trustworthy RAS initially have been identified. RAS must be (i) safe in any uncertain and dynamic surrounding environments; (ii) secure, thus protecting itself from any cyberthreats; (iii) healthy with fault tolerance; (iv) trusted and easy to use to allow effective human-machine interaction (HMI), and (v) compliant with the law and ethical expectations. Then, the challenges in implementing trustworthy autonomous system are analytically reviewed, in respects of the five key properties, and the roles of AI technologies have been explored to ensure the trustiness of RAS with respects to safety, security, health and HMI, while reflecting the requirements of ethics in the design of RAS. While applications of RAS have mainly focused on performance and productivity, the risks posed by advanced AI in RAS have not received sufficient scientific attention. Hence, a new acceptance model of RAS is provided, as a framework for requirements to human-centered AI and for implementing trustworthy RAS by design. This approach promotes human-level intelligence to augments human's capacity while focusing on contributions to humanity.

A new semantic attribute deep learning with a linguistic attribute hierarchy for spam detection

2017 International Joint Conference on Neural Networks (IJCNN), 2017

The massive increase of spam is posing a very serious threat to email and SMS, which have become ... more The massive increase of spam is posing a very serious threat to email and SMS, which have become an important means of communication. Not only do spams annoy users, but they also become a security threat. Machine learning techniques have been widely used for spam detection. In this paper, we propose another form of deep learning, a linguistic attribute hierarchy, embedded with linguistic decision trees, for spam detection, and examine the effect of semantic attributes on the spam detection, represented by the linguistic attribute hierarchy. A case study on the SMS message database from the UCI machine learning repository has shown that a linguistic attribute hierarchy embedded with linguistic decision trees provides a transparent approach to in-depth analysing attribute impact on spam detection. This approach can not only efficiently tackle 'curse of dimensionality' in spam detection with massive attributes, but also improve the performance of spam detection when the semantic attributes are constructed to a proper hierarchy.

Learnability and robustness of shallow neural networks learned by a performance-driven BP and a variant of PSO for edge decision-making

Neural Computing and Applications, 2021

With the advances in high-performance computing and graphics processing units, deep neural networ... more With the advances in high-performance computing and graphics processing units, deep neural networks attract more attention from researchers, especially in the areas of the image processing and natural language processing. However, in many cases, the computing resources are limited, especially in the edge devices of IoT enabled systems. It may not be easy to implement complex AI models in edge devices. The Universal Approximation Theorem states that a shallow neural network (SNN) can represent any nonlinear function. However, how fat is an SNN enough to solve a nonlinear decision-making problem in edge devices? In this paper, we focus on the learnability and robustness of SNNs, obtained by a greedy tight force heuristic algorithm (performance driven BP) and a loose force meta-heuristic algorithm (a variant of PSO). Two groups of experiments are conducted to examine the learnability and the robustness of SNNs with Sigmoid activation, learned/optimised by KPI-PDBPs and KPI-VPSOs, where, KPIs (key performance indicators: error (ERR), accuracy (ACC) and F 1 score) are the objectives, driving the searching process. An incremental approach is applied to examine the impact of hidden neuron numbers on the performance of SNNs, learned/optimised by KPI-PDBPs and KPI-VPSOs. From the engineering prospective, all sensors are well justified for a specific task. Hence, all sensor

Share price prediction of aerospace relevant companies with recurrent neural networks based on PCA

Expert Systems with Applications, 2021

The capital market plays a vital role in marketing operations for the rapid development of the ae... more The capital market plays a vital role in marketing operations for the rapid development of the aerospace industry. However, due to the uncertainty and complexity of the stock market and many cyclical factors, the stock prices of listed aerospace companies fluctuate significantly. This makes the share price prediction challengeable. To improve the prediction of share price for aerospace industry sector and well understand the impact of various indicators on stock prices, we provided a hybrid prediction model by the combination of Principal Component Analysis (PCA) and Recurrent Neural Networks. We investigated two types of aerospace industries (manufacturer and operator). The experimental results show that PCA could improve both accuracy and efficiency of prediction. Various factors could influence the performance of prediction models, such as finance data, extracted features, optimisation algorithms, and parameters of the prediction model. The selection of features may depend on the stability of historical data: technical features could be the first option when the share price is stable, whereas fundamental features could be better when the share price has high fluctuation. The delays of RNN also depend on the stability of historical data for different types of companies. It would be more accurate through using short-term historical data for aerospace manufacturers, whereas using long-term historical data for aerospace operating airlines. The developed model could be an intelligent agent in an automatic stock prediction system, with which, the financial industry could make a prompt decision for their economic strategies and business activities in terms of predicted future share price, thus

How Good a Shallow Neural Network Is for Solving Non-linear Decision Making Problems

Lecture Notes in Computer Science, 2018

The universe approximate theorem states that a shadow neural network (one hidden layer) can repre... more The universe approximate theorem states that a shadow neural network (one hidden layer) can represent any non-linear function. In this paper, we aim at examining how good a shadow neural network is for solving non-linear decision making problems. We proposed a performance driven incremental approach to searching the best shadow neural network for decision making, given a data set. The experimental results on the two benchmark data sets, Breast Cancer in Wisconsin and SMS Spams, demonstrate the correction of universe approximate theorem, and show that the number of hidden neurons, taking about the half of input number, is good enough to represent the function from data. It is shown that the performance driven BP learning is faster than the error-driven BP learning, and that the performance of the SNN obtained by the former is not worse than that of the SNN obtained by the latter. This indicates that when learning a neural network with the BP algorithm, the performance reaches a certain value quickly, but the error may still keep reducing. The performance of the SNNs for the two databases is comparable to or better than that of the optimal linguistic attribute hierarchy, obtained by a genetic algorithm in wrapper or in terms of semantics manually, which is much time-consuming.

Complexity of combinatorial ordering genetic algorithms COFFGA and CONFGA

THE 7TH INTERNATIONAL CONFERENCE ON APPLIED SCIENCE AND TECHNOLOGY (ICAST 2019), 2019

This paper analyses the complexity of two Algorithms called COFFGA (Combinatorial Ordering First ... more This paper analyses the complexity of two Algorithms called COFFGA (Combinatorial Ordering First Fit Genetic Algorithm) and CONFGA (Combinatorial Ordering Next Fit Genetic Algorithm). It also identifies the parameters that affect the performance of these algorithms. The complexity of the GA depends on the problem being solved by this GA, as well as the operators of the GA itself. The complexity of COFFGA and CONFGA are analysed individually. Even of these algorithms are slightly different, they may have extremely different complexities depending on the differences in their fitness function or termination condition. To provide a provable bound on a problem, there must be a bound on the evaluation function as well as a manner by which the underlying problem is tied to the representation. Given that there is no standard complexity of the GA, and the complexity of any GA depends on the problem that being solved by this GA and its operators, then CONFGA and COFFGA are analysed with different complexities; although they built upon the same algorithm and they are used to solve the same problem (Cloud resource allocation problem), but they are different in their operators their fitness function and termination condition.

A comparison of resource allocation process in grid and cloud technologies

Journal of Physics: Conference Series, 2018

Grid Computing and Cloud Computing are two different technologies that have emerged to validate t... more Grid Computing and Cloud Computing are two different technologies that have emerged to validate the long-held dream of computing as utilities which led to an important revolution in IT industry. These technologies came with several challenges in terms of middleware, programming model, resources management and business models. These challenges are seriously considered by Distributed System research. Resources allocation is a key challenge in both technologies as it causes the possible resource wastage and service degradation. This paper is addressing a comprehensive study of the resources allocation processes in both technologies. It provides the researchers with an in-depth understanding of all resources allocation related aspects and associative challenges , including: load balancing, performance, energy consumption, scheduling algorithms, resources consolidation and migration. The comparison also contributes an informal definition of the Cloud resource allocation process. Resources in the Cloud are being shared by all users in a time and space sharing manner, in contrast to dedicated resources that governed by a queuing system in Grid resource management. Cloud Resource allocation suffers from extra challenges abbreviated by achieving good load balancing and making right consolidation decision.