AIRG researchers conduct extensive AI focused research in the following areas:
- Knowledge Representation and Reasoning
- Intelligent Agents
- Applied Machine Learning
- Generative AI
- Game Theory
- Algorithmic Trading
- Reasoning with Large Language Models (LLMs)
- Multi-robotic Systems
- Intelligent Cybersecurity
Project title: Consequence Driven Solving and Hard SAT Instances.
Project description: It is known that the satisfiability problem (SAT) is NP-complete so that all NP problems are reducible to SAT and while SAT itself is in NP. As such, SAT is one of the important tools used in understanding the NP class of problems. In this work, we provide further insight on some structural aspects of SAT formulas that reveals what really makes them intractable in computational complexity. We further provide an algorithm that leverages this revealed structural properties and shows some promising benchmark results on random 3-SAT instances.
Contact: Dr Vernon Asuncion and Professor Yan Zhang
Project title: Automated Negotiation between Autonomous Vehicles
The emergence of self-driving cars has raised a significant research challenge for AI: How can autonomous vehicles adapt to the dynamics of road regulation and unforeseen traffic situations all by themselves? This project will deliver principles and mechanisms for automated negotiation among autonomous vehicles by building upon existing techniques from General Game Playing (GGP). We will extend the existing game description language (GDL) to support the description of arbitrary dynamic and unforeseen traffic situations for automatic processing by control systems for autonomous vehicles and develop mechanisms for autonomous vehicles to negotiate travel agreements when they meet on a road or at an intersection and encounter unforeseen, dynamic situations.
Contact: Associate Professor Dongmo Zhang
Project title: Power Quality Disturbance Feature Extraction and Recognition
Project description: Poor quality of power supplies could have the potential to interfere with communication networks, increase power losses, reduce life periods of electrical/electronic equipment, and cause a variety of malfunctions in power generation, transmission, distribution, and in end-users’ systems. Therefore, it is imperative to ascertain what power quality (PQ) problems the electricity grids are currently suffering and what are the formats and occurring frequencies of them, and then find out necessary countermeasures to mitigate the impacts they have been bringing about. Apparently, techniques of effective feature extraction and accurate classification are essential for the PQ disturbance recognition required by a smart grid. In the paper, after comparing some main feature extraction approaches, the authors present a PQ disturbance recognition scheme based on the combination of support vector machines and error correcting output codes. With the proposed feature extraction using Fourier and wavelet transforms respectively, the performance of the designed recognition system is verified. Simulations have shown that the proposed recognition methods, in particular, when using the Fourier transform, can achieve superior performance in terms of simplicity of feature extraction and high accuracy of the classification. Currently, the research is carried out to recognise more complicate non-stationary and concurrent multiple PQ disturbances.
Contact: Dr Jiansheng Huang
Project title: Smart Traffic Control for the Era of Autonomous Driving
Project description: Over the last decade, the research on autonomous vehicles (AVs) has made revolutionary progress, which brings us hope of safer, more convenient, and more efficient means of transportation. Most significantly, the advance of artificial intelligence (AI), especially machine learning, allows a self-driving car to learn and adapt to complex road situations with millions of accumulated driving hours, which are way higher than any experienced human driver can reach. However, autonomous vehicles on roads also introduce new challenges to traffic management, especially when we allow them to travel mixed with human driving vehicles. New theories for better understanding of the new era of transportation and new technologies for smart roadside infrastructures and intelligent traffic control are crucial for development and deployment of autonomous vehicles. This project aims to address these challenges, especially the social aspects of autonomous driving, including interaction between autonomous vehicles and roadside infrastructures, mechanisms of traffic management, the price of anarchy in road networks.
Contact: Associate Dongmo Zhang
Project title: Automatic Prompts Derivation Using Answer Set Programming
Using generative AI systems such as Midjourney, Stable Diffusion, DALL-E, etc., for image generations, have a wide range of applications. One important challenge in this task is to provide a sequence of precise prompts for the underlying requested image generation, so that proper images can be eventually generated meeting the user’s need. In this project, we develop an effective reasoning engine for prompts derivation, that can automatically associate relevant and non-redundant prompts together to generate specific features embedding in the output images.
Contact: Professor Yan Zhang and Dr Vernon Asuncion
Project title: Stable Diffusion-based Approach for Image Generation
Stable Diffusion (SD) is a current state-of-the-art approach for image generation. The main advantage of SD over generative adversarial networks (GANs) is that SD is more stable to train. This project aims to use this current SD technology to generate images in several domains that includes mountain landscape images and complex geometric fractal shapes.
Contact: Professor Yan Zhang
Project title: Charging Infrastructure Planning and Resource Allocation for Electric Vehicles
With the increasing uptake of electric vehicles (EVs) and relative lag in the development of charging facilities, how to plan charging infrastructure and electively use existing charging resources have become the top priority for governments, related industry and research communities. This project aims to address two key issues related to EV charging – charging station planning and charging resource allocation.
Contact: Associate Professor Dongmo Zhang
Project title: Image Aesthetics Assessment Using Deep Convolutional Neural Networks
Project description: Artificial Neural Networks (ANNs) is a machine learning model initially inspired by our own biological brain’s structure and function. In recent times, ANNs has become successful in the classification, clustering, pattern recognition, prediction and generation problem domains. In this project, we aim to use structured ANNs to classify the Image Aesthetic Assessment (IAA) as used in the domain of professional photography.
Contact: Dr Vernon Asuncion and Professor Yan Zhang
Project title: Designing new Quantum Machine Learning (QML) Algorithms
The project explores the intersection of quantum computing and machine learning to develop QML algorithms that can demonstrate advantages over classical counterparts. Specifically, this project will formulate quantum algorithms for certain machine learning tasks, and use quantum simulators such as IBM Qiskit to test and refine the algorithms. Experiments will be conducted to compare the efficiency and accuracy of QML algorithms with classical algorithms. Insights will be obtained on the practicality of QML algorithms for real-world applications.
Contact: Dr Weisheng Si