Research Areas

With the development of low-cost and low-power transceivers, sensors, and embedded processors, there has been growing interest in using sensor networks in a wide variety of applications, from surveillance and security to health and environmental hazard monitoring. In most applications, inference is usually the main task, and hence, the intelligent fusion of sensor information from geographically dispersed sensor nodes becomes an important issue. This project addresses some fundamental questions concerning the design and operation of sensor networks, and aims at developing theories that allow us to design and operate efficiently certain classes of sensor networks. The main research aims are to develop and contribute to a fundamental understanding of the use of feedback for intelligent data fusion in a sensor network​ and to propose signal processing and statistical inference methods for information fusion when perfect knowledge of the underlying distributions is not available.

We study randomized information dissemination methods in vehicular networks where the topology is dynamically varying. For example, information about an accident can be propagated to all vehicles using vehicle-to-vehicle wireless communications. The vehicles closest to the accident site have the most updated information, and this is communicated via a randomized gossiping procedure to all other vehicles within a certain range. We characterize the time required for a piece of information to be disseminated throughout the whole network, taking into consideration the changing topology of the network. 

In this project, we develop distributed optimization techniques for sensor networks in which the node-to-node communication is noisy. We consider objective functions that are separable, with each component known only locally at each sensor. The sensors exchange information with neighboring nodes (i.e., nodes within communication range) so that the network eventually finds a solution that is approximately optimal. We develop algorithms based on sub-gradient information exchange, and find conditions under which these algorithms converge.   

Much research effort has been focused on modelling the spread of rumours and viruses in a network, but identifying the information or “infection” sources remains a rather unexplored and challenging problem. In an increasingly connected world, identification of source nodes is expected to play an ever more important role in speedily uncovering the origins of a virus in a population, a rumour in a social network, or  the influential thought leaders in a network. This project aims to design algorithms that identify multiple infection sources in a network, based on knowledge of which nodes have the information, and the underlying network architecture.

This research thrust aims to develop fundamental theories of Wireless Sensor Networks (WSN) with a variety of applications including multi-agent cooperation, target detection, classification, tracking and self-organisation. Research activities address the challenges of the limited sensing and networking resources. Main research issues include the uncertainty modeling in unreliable networked environment, energy efficient quantization/ compression for bandwidth assignment, QoS (Quality of Service) provisioning via cross-layer integration, target detection, localization, tracking and inception, cooperation of multiple sensing agents, etc. Prototyping development is carried out for solutions to the real world WSNs application by developing wireless sensor nodes and test beds for multi-target tracking and inception, multi-formation control, body sensor network and networked air quality monitoring with limited energy, sensing, computing and wireless networking constraints.  

To build a human-like autonomous system that is capable of reasoning and decision making, it is imperative to encode knowledge into the system. Knowledge exists in various sources, but mostly in text and images, which are unstructured and are not machine understandable. This research aims to build an automatic knowledge extraction system that is able to extract knowledge from unstructured sources and then encode the knowledge extracted into machine understandable and usable form.

This research thrust aims at the development of advanced navigation, mapping, and localization of multi-vehicle surface and underwater vehicles. There are currently several collaborative projects with international partners underway. Autonomous surface craft (ASC) and autonomous underwater vehicle (AUV) localization and map estimation are extant research areas which lie at the core of any coast or ocean vehicular applications. The motivation behind the proposed research is to enable surface and subsea vehicles to interactively map and monitor the ocean environment. Recorded sensory information of the state of the ocean environment must be accompanied by reliable and consistent navigational data, providing information on exactly where data was recorded, and the geographic location of objects in the vicinity of the measurements.

This research thrust aims to develop traffic modelling techniques, regulation methodologies and simulation platforms to improve traffic flow efficiency and reduce fuel consumptions and environment pollution in megacities. With international partnerships, multiple projects are ongoing on the traffic flow modelling and predictions, traffic control techniques, road pricing/toll policies. These researches w​ill help government and relevant departments to form policy in urban traffic management and develop a new generation of traffic light control technologies. 

This research thrust focuses on the mobility, navigation and localisation of autonomous vehicles in urban environments. Key research areas include motion modelling and control, environment mapping, localisation, vision-based perception and identification, and sensor fusions. The applications include both outdoor and indoor semi-structured and unstructured environment. Research and development deliver both fundamental theories and new sensor designs. The group has developed many vehicle platforms with a variety of actuation configuration and sensor suites. Research results and expertise have been integrated into various robotic platforms for robotics competitions, including TechX Challenge organised by DSTA in Singapore and the World Robotcup competition.

Active Air Terminals (AAT) are popular in Europe, North America, and Australia etc. and have been widely used in many modern buildings. The interest for such technology is being fueled by their benefits including; energy saving, cost reduction, indoor environment quality improvement, reduced space requirements and so on. However, some technical challenges hindered the acceptance of the technology in tropical regions. In this project, a series of relevant technologies have been developed by the program team which will deliver the systematic solutions with design tools consisting of reliable design data, calibrated simulation models, design guidelines and software and methodologies for mechanical, control and management system implementations.

Deliver the exact of air through duct to the air-conditioned space is very important for ACMV systems in term of both energy saving and indoor environment quality. During system commissioning, TAB (Testing, Adjusting and Balancing) must be performed to satisfy the air distribution requirements. ​Current TAB is costly and time-consuming due to its iterative procedure. In this project, a non-iterative air TAB approach will be developed. New technologies will be integrated to solve the air balancing problems, including data mining, machine learning, optimal control and etc. The new TAB method can achieve air balancing automatically with the optimal control to minimize the fan power. The air balancing is also dynamic in responding to the changing air flow demand.

A water cooled central chiller plant is usually used to cool the space within a large building. It typically contributes 40-60% of the total energy consumption of the building, which is several times the plant capital cost during its lifetime. In order to verify the system performance effectively and ensure energy savings are achieved, a reliable measurement system that continuously tracks plant behaviour and plant performance is required. A rigorous method in measuring and verifying energy savings could lead to operational improvement and is instrumental in behavioral change programs, which could potentially lead to more significant energy savings. In this project, we will develop M&V related technologies for Singapore, and identify requirements to stipulate in M&V contract to ensure the set-up of robust M&V system for accurate verification of energy savings in commercial buildings with minimal additional costs.

Presently, the majority of dehumidification for building air-conditioning systems in Singapore is based on conventional mechanical dehumidification. In order to remove the moisture from the supply air, the process air should be cooled below its dew-point temperature, usually 12˚C, and reheating is needed to ensure comfortable supply air temperature. The cooling and reheating process wastes a large amount of electricity and suffers from high environmental cost. Moreover, the cooling coil works under the wet condition, which may cause health problem since condensed water makes the coil surface a breeding ground for bacterial. On the other hand, Liquid Desiccant Dehumidification System (LDDS) is considered as an attractive alternative to the conventional dehumidification technology. It uses the vapor difference between process air and liquid desiccant as the driven force to realize heat and mass transfer, consequently, dehumidifying the process air. Based on our patented LDDS technologies, this research project mainly focuses on advanced control strategies and avoiding liquid desiccant solution carrier over which will affect the Indoor Environmental Quality (IEQ) and occupants’ health besides corroding the ducting system and metal components as well.

In this project, a novel absorption refrigeration system consisting of horizontal type bubble absorber combining with air cooled heat exchanger. As the collapse of a bubble depends on higher heat and mass transfer effectiveness, the new will provide higher effectiveness to the absorption process. Availability of higher refrigerant mass ratio in strong solution provides a suitable platform to release higher amount of vapor refrigerant within the generator and increase the refrigeration capacity. Application of waste heat from industrial buildings and availability of solar thermal energy provides lower primary energy requirement to the system. The main objectives of the research are to optimize the existing horizontal bubble absorber to have higher absorption efficiency, optimize the generator to have higher generation effectiveness and to find the optimized to conditions to operate the system with highest overall efficiency.

In this project, BSTAR which is an Integrated Intelligent Building Management System (I2BMS) will be developed. BSTAR is a unique project development platform and management system for building and facility management, which integrate the function modules of monitoring, control, optimization, sensor automatic calibration, and fault detection and diagnosis (FDD). The platform retrieves from our inter-related projects and their inter-connections, embedding our breakthrough technologies such as cooling by low grade heat sources, Active Air Terminals (AAT), renewable energy based desiccant dehumidification, sensor network, Indoor Environmental Quality (IEQ) monitoring and control, visual analytics, and advanced cooperative control and optimization for building ACMV applications. These techniques will make the buildings more cost-effective, and will fill the gap between industry and academy for ACMC applications.