Large Language Model Assisted Dialogue for Cross-Domain IoT Collaboration
Shiyao Zhang
The PhD project titled “Large Language Model Assisted Dialogue for Cross-Domain IoT Collaboration” focuses on developing advanced dialogue systems that can facilitate effective communication and collaboration between Internet of Things (IoT) devices across different domains. The project aims to create a framework that leverages large language models to enable natural and intelligent dialogue interactions between IoT devices, thereby enhancing their ability to work together and provide more comprehensive and seamless services to users.
The research will investigate the interoperability challenges associated with cross-domain collaboration among IoT devices and enable collaborations with multiagent dialogue systems under such conditions. The project will explore the use of large language models, which have captured knowledge that is not guaranteed to be the same as humans-created, to understand and generate natural language dialogue, which can bridge the differences and enable effective communication between devices from different domains, to better fit large language model in the loop.
The project will also investigate the use of reinforcement learning techniques to improve the dialogue policies of IoT devices, allowing them to learn from their interactions and adapt their dialogue strategies to better fulfil requirements. Additionally, the research will aim to ensure the security and privacy of the dialogue interactions, addressing potential vulnerabilities and threats that may arise from the use of large language models in IoT collaboration.
Overall, the PhD project aims to contribute to the advancement of IoT technology by creating a large language model-assisted dialogue framework that enables cross-domain collaboration among IoT devices, leading to more intelligent, user-centric, and efficient IoT ecosystems.
Collaborative Reinforcement Learning for Active SLAM
Xueyan Yao
The PhD project titled “Collaborative Reinforcement Learning for Active SLAM” focuses on the development of collaborative simultaneous localization and mapping (SLAM) systems for multiple robots using reinforcement learning techniques. The project aims to create a framework that allows multiple robots to work together efficiently in unknown environments, improving their ability to build accurate maps and localize themselves while minimizing the time and resources required.
The research will investigate the challenges associated with collaborative SLAM, such as communication constraints, sensor limitations, and dynamic environments. The project will explore the use of reinforcement learning, an artificial intelligence technique that enables agents to learn optimal policies through interactions with their environment, to develop collaborative SLAM algorithms that can adapt and learn from experience.
The project will also investigate the use of active learning techniques, which involve selecting the most informative actions to reduce uncertainty, to improve the performance of the collaborative SLAM system. By leveraging the computational power of multiple robots and allowing them to share information and learn from each other’s experiences, the research aims to create a more efficient and robust SLAM system.
Overall, the PhD project aims to contribute to the advancement of robotics and autonomous systems by developing a collaborative reinforcement learning framework for active SLAM, which can have wide-ranging applications in areas such as search and rescue missions, autonomous navigation, and smart environments.
Multi-sensory Interactions under Cross-device Collaborative IoT Environments
Yue Li
The PhD project titled “Multi-sensory Interactions under Cross-device Collaborative IoT Environments” approaches the intersection of human-computer interaction (HCI) and the Internet of Things (IoT) from a unique perspective. This research is centered on understanding how humans perceive and interact with IoT devices that collaborate across different platforms. The project’s focus is on designing interactions that engage multiple human senses, such as touch, sight, and sound, to create more intuitive and immersive experiences.
The research will delve into the challenges of designing interfaces and interactions that effectively utilize cross-device collaboration in IoT systems. It will explore how sensory information can be integrated and presented to users in a way that enhances their understanding and control of their environments. The project will also investigate the role of sensory feedback in improving the usability and user experience of IoT applications.
By conducting human studies and user-centered design processes, the research aims to identify the most effective multi-sensory interaction patterns and design principles. This will involve iterative prototyping, user testing, and evaluation to ensure that the developed interfaces are both technically feasible and align with human cognitive and sensory processes.
Ultimately, the PhD project seeks to contribute to the field of HCI by providing insights into how multi-sensory interactions can be leveraged to create more natural and engaging user experiences in cross-device collaborative IoT environments. The outcomes of this research are expected to inform the design of future IoT systems, making them more accessible and satisfying for users in various contexts, including smart homes, wearable technology, and public spaces.
Blockchain-empowered Trustworthy IoT Data Management and Trading
Sida Huang
The PhD project titled “Blockchain-empowered Trustworthy IoT Data Management and Trading” aims to develop a robust framework for managing and trading IoT (Internet of Things) data using blockchain technology. The project focuses on addressing the challenges associated with data privacy, security, and trust in IoT ecosystems.
By leveraging the decentralized and immutable nature of blockchain, the project seeks to create a secure and transparent approach for IoT devices to share and monetize data. The research will explore the potential of blockchain technology to ensure the integrity and non-repudiation of IoT data transactions.
The project will also investigate the use of blockchain to enforce access control and data ownership policies, allowing IoT device owners to maintain control over their data and determine who can access it. Additionally, the research will aim to optimize the performance and scalability of blockchain-based IoT data management and trading systems to accommodate the massive scale of IoT devices and data.
Overall, the PhD project aims to contribute to the advancement of IoT technology by creating a trustworthy and efficient framework for managing and trading IoT data, which can have wide-ranging applications in smart cities, healthcare, supply chain management, and more.
Federated Learning-Driven Edge Resource Allocation for IoT Applications
Medhav Kumar Goonjur
Artificial intelligence (AI) has ushered in a new era of technological advancement, profoundly impacting various domains within the Internet of Things (IoT), including smart cities, autonomous vehicles, predictive maintenance, and advanced healthcare systems. To achieve an efficient IoT ecosystem, it is imperative to judiciously allocate communication and computational resources, addressing the inherent latency and energy consumption challenges. As the IoT proliferates, devices like smartwatches, equipped with sophisticated sensors, are capable of collecting, processing, and storing vast amounts of data. However, processing this deluge of raw data within a conventional centralized machine learning (ML) framework is impractical, due to the prohibitive costs and inefficiencies associated with transmitting and processing data at a centralized cloud server. Moreover, it raises significant concerns regarding data privacy.
To alleviate these challenges, a distributed computing approach known as Mobile Edge Computing (MEC) has emerged. This model deploys servers at the network’s edge nodes, facilitating the execution of computationally intensive tasks and significantly reducing latency. MEC operates on a three-tiered architecture: the user equipment layer, the MEC layer, and the cloud layer.
Federated Learning (FL), a novel AI paradigm, addresses data privacy concerns by allowing the global model from the central server to be distributed as local models at each edge node. These local models transmit their parameters to the central server for aggregation and global model training. Once trained, the updated global model parameters are disseminated to the local models for optimization in subsequent iterations, ensuring that the model converges to the desired accuracy. Nevertheless, the efficient allocation of resources, such as computation and communication, across these resource-constrained devices remains a significant challenge.
The PhD project’s primary focus is on leveraging FL to optimize resource allocation within an MEC framework for IoT applications. The objectives are as follows:
Integrate FL into the IoT ecosystem to bolster privacy-preserving mechanisms.
Employ deep reinforcement learning techniques to enhance resource allocation and energy efficiency within the IoT framework.
Evaluate the proposed framework in both simulated and real-world testbed environments.
Develop an innovative framework that maximizes resource utilization while safeguarding user privacy within an edge-intelligent IoT framework.
Doctoral Project
Large Language Model Assisted Dialogue for Cross-Domain IoT Collaboration
Shiyao Zhang
The PhD project titled “Large Language Model Assisted Dialogue for Cross-Domain IoT Collaboration” focuses on developing advanced dialogue systems that can facilitate effective communication and collaboration between Internet of Things (IoT) devices across different domains. The project aims to create a framework that leverages large language models to enable natural and intelligent dialogue interactions between IoT devices, thereby enhancing their ability to work together and provide more comprehensive and seamless services to users.
The research will investigate the interoperability challenges associated with cross-domain collaboration among IoT devices and enable collaborations with multiagent dialogue systems under such conditions. The project will explore the use of large language models, which have captured knowledge that is not guaranteed to be the same as humans-created, to understand and generate natural language dialogue, which can bridge the differences and enable effective communication between devices from different domains, to better fit large language model in the loop.
The project will also investigate the use of reinforcement learning techniques to improve the dialogue policies of IoT devices, allowing them to learn from their interactions and adapt their dialogue strategies to better fulfil requirements. Additionally, the research will aim to ensure the security and privacy of the dialogue interactions, addressing potential vulnerabilities and threats that may arise from the use of large language models in IoT collaboration.
Overall, the PhD project aims to contribute to the advancement of IoT technology by creating a large language model-assisted dialogue framework that enables cross-domain collaboration among IoT devices, leading to more intelligent, user-centric, and efficient IoT ecosystems.
Collaborative Reinforcement Learning for Active SLAM
Xueyan Yao
The PhD project titled “Collaborative Reinforcement Learning for Active SLAM” focuses on the development of collaborative simultaneous localization and mapping (SLAM) systems for multiple robots using reinforcement learning techniques. The project aims to create a framework that allows multiple robots to work together efficiently in unknown environments, improving their ability to build accurate maps and localize themselves while minimizing the time and resources required.
The research will investigate the challenges associated with collaborative SLAM, such as communication constraints, sensor limitations, and dynamic environments. The project will explore the use of reinforcement learning, an artificial intelligence technique that enables agents to learn optimal policies through interactions with their environment, to develop collaborative SLAM algorithms that can adapt and learn from experience.
The project will also investigate the use of active learning techniques, which involve selecting the most informative actions to reduce uncertainty, to improve the performance of the collaborative SLAM system. By leveraging the computational power of multiple robots and allowing them to share information and learn from each other’s experiences, the research aims to create a more efficient and robust SLAM system.
Overall, the PhD project aims to contribute to the advancement of robotics and autonomous systems by developing a collaborative reinforcement learning framework for active SLAM, which can have wide-ranging applications in areas such as search and rescue missions, autonomous navigation, and smart environments.
Multi-sensory Interactions under Cross-device Collaborative IoT Environments
Yue Li
The PhD project titled “Multi-sensory Interactions under Cross-device Collaborative IoT Environments” approaches the intersection of human-computer interaction (HCI) and the Internet of Things (IoT) from a unique perspective. This research is centered on understanding how humans perceive and interact with IoT devices that collaborate across different platforms. The project’s focus is on designing interactions that engage multiple human senses, such as touch, sight, and sound, to create more intuitive and immersive experiences.
The research will delve into the challenges of designing interfaces and interactions that effectively utilize cross-device collaboration in IoT systems. It will explore how sensory information can be integrated and presented to users in a way that enhances their understanding and control of their environments. The project will also investigate the role of sensory feedback in improving the usability and user experience of IoT applications.
By conducting human studies and user-centered design processes, the research aims to identify the most effective multi-sensory interaction patterns and design principles. This will involve iterative prototyping, user testing, and evaluation to ensure that the developed interfaces are both technically feasible and align with human cognitive and sensory processes.
Ultimately, the PhD project seeks to contribute to the field of HCI by providing insights into how multi-sensory interactions can be leveraged to create more natural and engaging user experiences in cross-device collaborative IoT environments. The outcomes of this research are expected to inform the design of future IoT systems, making them more accessible and satisfying for users in various contexts, including smart homes, wearable technology, and public spaces.
Blockchain-empowered Trustworthy IoT Data Management and Trading
Sida Huang
The PhD project titled “Blockchain-empowered Trustworthy IoT Data Management and Trading” aims to develop a robust framework for managing and trading IoT (Internet of Things) data using blockchain technology. The project focuses on addressing the challenges associated with data privacy, security, and trust in IoT ecosystems.
By leveraging the decentralized and immutable nature of blockchain, the project seeks to create a secure and transparent approach for IoT devices to share and monetize data. The research will explore the potential of blockchain technology to ensure the integrity and non-repudiation of IoT data transactions.
The project will also investigate the use of blockchain to enforce access control and data ownership policies, allowing IoT device owners to maintain control over their data and determine who can access it. Additionally, the research will aim to optimize the performance and scalability of blockchain-based IoT data management and trading systems to accommodate the massive scale of IoT devices and data.
Overall, the PhD project aims to contribute to the advancement of IoT technology by creating a trustworthy and efficient framework for managing and trading IoT data, which can have wide-ranging applications in smart cities, healthcare, supply chain management, and more.
Federated Learning-Driven Edge Resource Allocation for IoT Applications
Medhav Kumar Goonjur
Artificial intelligence (AI) has ushered in a new era of technological advancement, profoundly impacting various domains within the Internet of Things (IoT), including smart cities, autonomous vehicles, predictive maintenance, and advanced healthcare systems. To achieve an efficient IoT ecosystem, it is imperative to judiciously allocate communication and computational resources, addressing the inherent latency and energy consumption challenges. As the IoT proliferates, devices like smartwatches, equipped with sophisticated sensors, are capable of collecting, processing, and storing vast amounts of data. However, processing this deluge of raw data within a conventional centralized machine learning (ML) framework is impractical, due to the prohibitive costs and inefficiencies associated with transmitting and processing data at a centralized cloud server. Moreover, it raises significant concerns regarding data privacy.
To alleviate these challenges, a distributed computing approach known as Mobile Edge Computing (MEC) has emerged. This model deploys servers at the network’s edge nodes, facilitating the execution of computationally intensive tasks and significantly reducing latency. MEC operates on a three-tiered architecture: the user equipment layer, the MEC layer, and the cloud layer.
Federated Learning (FL), a novel AI paradigm, addresses data privacy concerns by allowing the global model from the central server to be distributed as local models at each edge node. These local models transmit their parameters to the central server for aggregation and global model training. Once trained, the updated global model parameters are disseminated to the local models for optimization in subsequent iterations, ensuring that the model converges to the desired accuracy. Nevertheless, the efficient allocation of resources, such as computation and communication, across these resource-constrained devices remains a significant challenge.
The PhD project’s primary focus is on leveraging FL to optimize resource allocation within an MEC framework for IoT applications. The objectives are as follows: