* Funded Research Projects
∞
What I had done and what I am doing might not necessarily be what are listed ∞
It aims to develop
an innovative solution for search and coverage to navigate through complicated
ocean environments. This approach leverages the active sensing capabilities of
multi-robot systems to supplement digital twin model, offering a more
comprehensive and real-time understanding of the environment. Based on point
cloud data, which are inherently non-convex and unstructured, this method
efficiently generates collision-free Voronoi regions using only local sensing
information through spatial decomposition and spherical mirroring techniques. A
deadlock-aware guided map integrated with a gradient-optimized and centroid
Voronoi-based coverage control policy is constructed to improve efficiency by
avoiding exhaustive searches and local sensing pitfalls. In addition, dynamic
connectivity maintenance method is designed, aiming to enhance multi-robot
coordination, maximize coverage efficiency, and hasten marine target detection.
Funding Source: University
Grants Committee, Hong Kong, 2025–2027 (with Patrick Chen and Xinyi Wang –
Co-Is)
In
this project, we propose an innovative collaborative task assignment of
multi-agent unmanned systems for general infrastructure inspection. The main
objective of the infrastructure inspection system with multiple UAVs or drones
is to design the best task assignment strategy and the most efficient paths
that enable a team of drones to automatically collect high-quality data.
Compared with the single-agent methods, the proposed approach has the
advantages of being more efficient and effective. It can also resolve task
complexity and enhance the overall system reliability. The proposed project
links the research on multi-agent systems to real industrial applications. It
can also be further integrated with advanced artificial intelligence and
digital twin techniques to develop an information management system within a
smart city framework.
Funding Source: University
Grants Committee, Hong Kong, 2024–2026
In
this project, we aim to develop an advanced search and pursuit-evasion strategy
that enable multi-agent systems, specifically UAVs or drones, to search and
track an evader cooperatively and autonomously in cluttered environments
without obstacle and inter-vehicle collisions. It aims to provide a standard
search and pursuit-evasion framework for general multi-agent unmanned systems.
The proposed research, once realized, is expected to play an instrumental role
for unmanned aerial systems in preventing safety incidents caused by intruding
agents from happening and thus avoiding casualties and property losses.
Funding Source: University
Grants Committee, Hong Kong, 2023–2025 (with Panpan Zhou – Co-I)
In
this project, we aim to develop advanced motion planning techniques that enable
multiple unmanned systems, specifically UAVs or drones, to autonomously
navigate and fly in cluttered environments without inter-vehicle collisions. It
aims to provide a standard motion planning framework for general multi-agent
unmanned systems. The proposed research, once realized, is expected to play an
instrumental role for unmanned aerial systems in numerous industrial
applications, such as reconnaissance for search and rescue, environment
monitoring, security surveillance, powerline and pipeline inspection, and
geographical mapping.
Funding Source: University
Grants Committee, Hong Kong, 2022–2024
In
this research project, we aim to study and investigate a robust and
sophisticated flight control system framework with intelligent motion planning
techniques using model predictive control (MPC) and a robust and perfect
tracking (RPT) control method for unmanned aerial vehicles. The MPC-based
motion planning is incorporated with an appropriate neural network and a
derivative-free optimization to generate optimal paths and trajectory
references for UAVs. This allows the UAV to be operated in real and complex
environments. The RPT control is used to perfectly and robustly track the
resulting reference, and yield good flight performance.
Funding Source: University
Grants Committee, Hong Kong, 2021–2023
Funding Source: TSSSU, Innovation
and Technology Commission, Hong Kong, 2024–2025 (with Patrick Chen ~ PI)
Funding Source: Environment and Conservation Fund, Hong Kong, 2024–2026
(with Patrick Chen ~ PI)
Funding Source: Electrical
and Mechanical Services Department, Hong Kong, 2024–2025 (with Patrick Chen
~ PI)
The
objective of this project is to use unmanned systems technology as a tool for
building inspection. It aims to develop an automated building inspection and
information system that can constantly monitor the structural health of
buildings. It includes (i) inspection and data collection using fully
autonomous drones equipped with necessary sensor, and (ii) an artificial
intelligence (AI) platform for data processing, which will eventually be
integrated to a BIM (building information management) system to generate
building structural health report.
Funding Source: Hong Kong Center for Logistics Robotics, Hong Kong,
2020–2025
In
this project, we aim to develop a hybrid aircraft that can be operated in both
air and underwater. Aerial-aquatic designs are an emerging type of multimodal
vehicles. Similar to the concept of amphibious vehicles and land-air hybrids,
aerial-aquatic vehicles expand the range of operation and application
possibilities of mobile robots by being able to operate in two mediums. The
uniqueness of aerial-aquatic vehicles has many potential applications. Issues
related to mechanical design, automatic control and underwater SLAM will be
investigated. The project also investigates other underwater unmanned systems.
Funding Source: Shanghai Research Institute for Intelligent Autonomous Systems,
Shanghai, 2019–
It aims to develop a software framework and software-in-the-loop
simulation as well as hardware-in-the-loop simulation platforms for general
autonomous unmanned systems, which include modules such as (i) sensing,
perception and localization, (ii) dynamic modeling, (iii) automatic control,
(iv) motion planning, (v) task planning and mission management, (vi) data
processing and analysis, and (vii) complete system integration.
Funding Source: Peng Cheng Laboratory, Shenzhen, 2019-2022
This project aims to develop a comprehensive and implementable GPS-less
navigation system for unmanned aerial vehicles. The system is expected to be
fully integrated on-board, light-weight, real-time, robust and without
unrealistic assumptions. To achieve the target, the following four problems
need to be investigated in depth, where innovation and optimization have to be
applied on both hardware and software aspects: (1) Multi-sensory data fusion; (2)
3D real-time simultaneous localization and mapping; (3) Dynamic 3D Path
planning in unknown environments; and (4) Multi-UAV cooperative control.
Funding Source: Defence Science &
Technology Agency, Singapore, 2015–2019, S$1,668,000
In this project, we aim to investigate the problem of intelligent
navigation of MAVs in realistic indoor and outdoor cluttered environments
without global referencing resources. We intend to achieve this target by
systematic integration of achievements of the following four tasks: (1) MAV
platform design with small size but sufficient payload and endurance; (2)
Robust and precise ego-motion estimation in indoor and outdoor cluttered
environments via onboard relative sensors; (3) Navigation in cluttered
environments with obstacle avoidance; and (4) Smooth navigation transition
between indoor and outdoor environments. The proposed algorithms and methods
will be tested and verified using actual MAV platforms. Multiple indoor and
outdoor scenarios will be defined or developed for the developed MAVs to
demonstrate its functionalities. We have also planned to demonstrate and test
the developed algorithms and platforms at international and local events.
Funding Source: DSO National Laboratories, Singapore, 2014–2016, S$492,000
This portion of work also requires the controlled UAV to demonstrate the
capability of visual servoing, provided that some natural
landmarks are within the view of the UAV onboard camera. During visual servoing, the controlled UAV should be able to perform
stable ascending, descending and yawing by visually tracking the selected
feature. In addition, the capabilities of the controlled UAV to perform
automatic perching on a selected rooftop location (without artificial landmark)
and landing on a moving platform (with man-made but discreet landmarks) are to
be developed. All these UAV capabilities should be demonstrated in a fully
autonomous fashion and without using any GPS device.
Funding Source: Temasek Laboratories, National University of Singapore, 2014–2016, S$150,000
We propose a hierarchical approach utilizing the optimization methods,
graph theory approaches and image processing techniques to unify the planning
and control levels. The optimal planning algorithms equip the team of UAVs with
the distributed decision-making subject to spatial-temporal performance and
constraints. The information acquired from the distributed measurement and
optimal coverage flights allocate individual tasks for UAVs to be accomplished
by target/trajectory tracking algorithm in the real-time control layer. The
techniques proposed in this project are important in many applications, such as
air defense and urban application such as traffic control and crisis
management. For example, multiple UAVs work as a team for cover a wide area
with accurate image measurement and real-time control of the traffic flow.
Another example would be recognition and tracking of a suspect/enemy.
Funding Source: Temasek Defence
Systems Institute, 2013–2016, S$300,000 (with C Xiang ~ PI,
T H Lee, C Chen, W Kang and O. Yakimenko)
In this project, we propose to develop an advanced outdoor navigation
system to explore the theories and technologies that enable UAVs to realize
autonomous navigation in outdoor cluttered environments, especially forest. To
realize the navigation system, several main topics need to be investigated,
including advanced sensing technologies, sophisticated navigation approaches,
as well as simultaneous localization and mapping (SLAM) techniques. A variety
of sensing technologies are considered in the project, including
electro-optical sensors (EO), light detection and ranging sensors (LIDAR),
inertial measurement units (IMU), the global positioning system (GPS), and so
on. The fusion techniques are investigated to combine measurements of these
sensing technologies to realize obstacle detection and robust navigation even
at the loss of GPS signal. In addition, based on obstacle detection, dynamic
path planning scheme is studied to determinate a safe path for the UAV to
achieve required missions. Moreover, special attention is paid to the
simultaneous localization and mapping in largescale environments. It is
important to rigorously analyze the theories and techniques on how to integrate
a local map into a global map and on how to smoothly transform between these different
navigation conditions.
Funding Source: Temasek Defence Systems Institute, Singapore, 2012–2015, S$300,000 (with T H Lee, C Chen and O.
Yakimenko)
The DARPA UAVForge Challenge is a Defense
Advanced Research Projects Agency (DARPA) and Space and Naval Warfare Systems
Center Atlantic (SSC Atlantic) collaborative initiative to design, build and
manufacture advanced small unmanned air vehicle (UAV) systems. The NUS GremLion team has been selected as a finalist for the UAVForge Fly-Off competition. In order to achieve the goals
of the UAVForge, a navigation system for UAVs has to
be developed to cope with the challenges caused by the high demanding
requirements, such as long-range navigation, vehicle following, obstacle avoidance,
and rooftop landing, and the constraints of the UAV platform with limited
payload and energy.
The main tasks of this project involve the hardware construction of an
embedded vision aided navigation system, development and realization of the required
mission algorithms, and implementation of the proposed system on actual UAV
platforms. First, the vision aided navigation system is developed, which
includes multiple sensors, a vision processing unit, data and video links, etc.
Second, mission algorithms are investigated and implemented on the navigation
system. The mission algorithms consist of vision-aided obstacle detection and
avoidance, and vision-based autonomous rooftop landing and target following.
Finally, the proposed navigation system and mission algorithms will be tested
by using the UAV platform provided by DSO in real flight.
Funding Source: DSO National Laboratories, 2012, S$230,040
In this proposed project, we take the challenge to develop an
ultra-compact micro aerial vehicle (MAV) with about 100 g in mass and 8 minutes
flight time. The developed MAV is able to safely navigate through indoor
environment and complete autonomously necessary flight missions. The main tasks
of this project involve the construction of the custom-made aerial vehicles,
development of an embedded avionic system, modeling of the MAV system and
design of robust flight controllers. The systematic design methodology and
innovative technologies will be utilized to optimize the vehicle itself and the
avionic system to fulfill the required specifications. A robust flight
controller will be designed and implemented for the MAV in terms of the
identified model. The entire system of the MAV will be tested in the actual
flight.
Funding Source: DSO National Laboratories, Singapore, 2011–2013, S$625,800 (with T. H. Lee and P. Tan)
The main objective of this project is to develop theory and algorithms
of computational methods for optimal UAV trajectory planning in obstacle-rich
environment. The proposed research is to explore the application of the
state-of-the-art in computational optimal control to optimal UAV trajectory
planning in a challenging environment with obstacles, a problem that cannot be
satisfactorily solved using existing technologies. Autonomous UAV trajectory
planning that optimizes performance has significant potential in various
applications of military, anti-terrorist, and civilian operations.
Funding Source: Temasek Defence Systems Institute, National University of Singapore, 2010–2013, US$148,521 (with W Kang
from NPS ~ PI)
The recent success of unmanned aerial vehicles (UAVs) in the military
and civilian applications has brought great interests in developing new
generation of UAVs that are capable of flying fully autonomously in an unknown
environment, especially in an indoor environment. A UAV equipped with a
sophisticated indoor navigation system is expected to be an ideal platform for
a wide range of military and civilian applications. To realize such
applications, the indoor navigation system for UAVs has to be developed to cope
with the challenges caused by the complicated indoor environment (such as
scattered obstacles and denied reception of GPS signals) and the constraints of
the UAV platform (such as its instability and limited payload).
In this proposed project, we aim to develop a 3D indoor navigation
system, which is able to aid UAVs to safely navigate through the unknown and
complicated indoor environment and complete autonomously necessary flight
missions. The main tasks of this project involve hardware construction of an
embedded navigation system (including multiple sensors, data and image
processing units, and data and video links), the development and realization of
robust 3D indoor navigation algorithms, and the tests of the system on the
actual UAV platforms. In the proposed navigation scheme, sophisticated machine
vision algorithms such as robust feature extraction, an optical flow method and
stereo vision approach are utilized to stabilize the UAVs and avoid the
obstacles. The output of multiple sensors including inertial measurement
sensors, visual sensors, and range sensors are employed to realize the 3D
simultaneously localization and mapping (SLAM) for UAVs.
Funding Source: Temasek Defence Systems Institute, National University of Singapore, 2009–2012, S$300,000 (with H Lin and
T H Lee)
The success of using machine vision technologies for multi-agent system
(MAS) in the military and civilian applications has aroused great interests in
its potential in future. One of the leading-edge goals of the next generation
vision-based MAS aims to the capability of navigation in complex environments
such as indoor areas. Two challenges of this goal are the efficient and
reliable testbeds design and the advanced vision-based navigation system for
MAS to assist GPS-only or gyro-less navigation systems. So far, there has been
very little research related to this field and many core issues left to be
studied to support this application.
In this project, we will focus on two main issues: The first one is to
design and implement more sophisticated mechatronic system design methodology
to construct reliable testbeds of the multi-agent system (MAS), based on
off-the-shelf ground and aerial vehicles; the second part is to investigate an
advanced vision-based navigation system, aided by machine vision technologies,
to guide MAS in complex environments. For the former, a small fleet of
low-cost, ultra-light weight, but reliable MAS testbeds will be developed. For
the vision-based navigation system, advanced vision-augmented inertial or GPS
approaches will be employed to stabilize the MAS in GPS-only or gyro-less
environments. Vision-based motion coordination, target detection and tracking
algorithms will be designed for MAS, which fuse information of multiple
platforms, prior knowledge and human commands in diverse navigation modes.
Funding Source: Temasek Laboratories, National University of Singapore, 2009–2013, S$200,000 (with K Y Lum
and K Peng)
Cooperative control of multiple Unmanned Air Vehicles (UAVs) is still in
its infancy and poses significant theoretical and technical challenges.
Usually, the cooperative control scheme is organized in a hierarchy manner,
where the low-level control signals and the higher-level supervisory logical
rules are designed separately. This project aims to propose a new hybrid system
approach that serves as a unified framework to study the coupling between the
essentially discrete features of the cooperative supervisory logic and the
continuous dynamics of UAVs. This perspective allows us to develop new
techniques to determine the effects of the coupling in the performance of the
system, and, more importantly, innovative hybrid control technologies for the
cooperative reconfiguration control of multiple UAVs. It is believed that
significant benefits can potentially be obtained through a joint design of the
supervisory logic and the continuous control algorithms. The outcome of the
project will be on a novel hybrid cooperative control concepts, modeling and
design framework for UAV groups so as to achieve quick responses and autonomous
reconfiguration ability for different tasks or scenarios. Actual flight testing
will be conducted to verify the developed hybrid control technologies.
Funding Source: Temasek Defence Systems Institute, National University of Singapore, 2008–2011, S$300,000 (with H Lin ~
PI, T H Lee and C Chen)
UAVs aroused a great interest in past decades because of their military
and civil applications. An autonomous UAV can, however, carry out only certain
scheduled tasks and it has no capability of responding to unscheduled and
expected events. An operating platform is thus needed to enhance the capability
of UAVs. Because of the limitation of radio wave range, the platform should be
mobile. A manned air vehicle (MAV) is a good moving platform to lead UAVs in
flight formation. Flight formation of UAVs via an MAV has great advantages
compared to UAVs or MAVs in separation. It can make full use of either human
abilities or properties of UAVs to improve possibility and feasibility to complete
designated missions. So far, there is a little research related to this field
and much core issues left to be studied to support this application.
This project is to explore new concepts, definitions and technologies
applicable to lead UAVs via MAVs. We aim to develop a flight control platform
for formation flight and test the developed technologies and concepts using
actual UAV helicopters. In particular, we aim (1) to develop systematic
concepts and methods to model and control flight formation of UAVs via MAVs;
(2) to develop a flight control platform for formation flight; and (3) to
verify the developed technologies and concepts using actual UAV helicopters.
Funding Source: Temasek Defence
Systems Institute, 2007–2010, S$300,000 (with T H Lee and
Rodney Teo)
Flight formation comprising UAV helicopters and piloted helicopters has
become a hot topic because of its potential applications in both civil and
military domains. The challenges in designing flight control systems for UAV
helicopters and their flight formation are nonlinearities in the helicopter
aerodynamics and complexity associated with flight formation. Diverse control
techniques have been proposed in the literature to tackle such challenges. But
there is still much left to be further investigated.
In this project, we will focus on investigating identification
methodologies for complex systems and nonlinear system control methods for
solving the proposed problem. A systematic method incorporating the
Funding Source: Temasek Laboratories, National University of Singapore, 2007–2009, S$100,000 (with K Y Lum
and K Peng)
This project is to pursue advance nonlinear control methods to design
flight control systems for flying vehicles, namely a radio-controlled
helicopter and an ST Aerospace FanTail air vehicle.
The following topics will be focused in the researches: (1) to build a test bed
of the flight control systems; (2) to pursue systematic methods to model the
helicopter and FanTail with in-flight data and to
establish a base of the helicopter for the design of helicopter flight control
systems; and (3) to pursue and verify advance nonlinear control methods to
design helicopter and FanTail flight control systems
for super maneuver flights to complete the mission of terrain follow &
obstacle avoidance or attacks onto the targets on the ground.
The proposed research has both major potential theoretical impact and
engineering impact. It is to be focused at systematic methods to model the
helicopters and FanTail with in-flight data and
nonlinear control methods to design integrated flight control systems to improve
the ability of the helicopter and FanTail to complete
the designated missions. These methods will be verified on the test bed to be
built, a radio-controlled helicopter and FanTail. The
proposed methods will be significantly useful to improve the level of the
design of both military and civil helicopters as well as other flying vehicles.
Funding Source: Defence Science & Technology
Agency, Singapore, 2003–2007, S$700,000
The friction in the rotary actuator pivot bearing deteriorates
significantly the performances of hard disk drive servo systems especially as
the actuator inertia drops in small hard disk drives. The residual errors
caused by the friction make head positioning servo systems difficult to
maintain the read/write head over the narrower track center. The mitigation of
the friction is an ongoing issue studied by many researchers. It has become one
of challenges to design hard disk drive servo systems for small hard disk
drives.
This research project focuses on the issues associated with the
compensation of the friction in hard disk drives. More specifically, we will
(1) develop a systematic method to model all the features of the friction in
the actuator rotary pivot bearing; (2) develop compensation algorithms to
implement the mitigation of the friction; and (3) investigate the improvements
in access time, residual bias and positioning error signal by the mitigation of
the friction through implementations.
Funding Source: University Academic Research, National University of Singapore, 2003–2006, S$130,432 (with T H Lee)
In this project, we propose to develop a virtual reality remote user
interface for the web-based remote laboratories that we have successfully
launched in the Department of Electrical and Computer Engineering, the National
University of Singapore. The system can be used for distance education through SingAREN, a broadband high-speed data communication
network. We will develop real-time applications for the system, and utilize the
SingAREN as the backbone for the high-speed data
communication needed to transmit high quality 3D graphics for the web-based
remote laboratories. The use of SingAREN will be
crucial to the project as conducting real-time remote experiments requires
instantaneous AV feedback and communications. The project is an R&D effort
that falls within the scope of distant e-education which will become more and
more important in the increasingly connected global digital world.
The objectives of the project are as follows: (1) to design novel
techniques and processing systems for 3D virtual reality in the context of
web-based experimentation; (2) to evaluate the techniques, algorithms and
protocols developed from a AV feedback and communication point of view; (3) to
integrate a 3D virtual reality system for the web-based experiments that have been
developed in the Department of Electrical and Computer Engineering, the
National University of Singapore; and (4) to develop a set of standard routines
in a testbed through which a 3D virtual reality can be readily adopted for
remote web experimentation.
Funding Source: Singapore Advanced Research &
Education Network, Singapore, 2001–2003, S$336,500 (with C C Ko ~ PI)
Due to the phenomenal increase in the recording density in hard disk
drives (HDD), the requirement on the HDD head positioning servo system is
escalating. Because of the narrower tracks at higher track densities, the servo
bandwidth must be increased in order to reject disturbances quickly and
maintain the read/write head over the track center. Some industry experts
predict that it will take only few years for the track density to reach 25,000
tracks per inch (TPI). At such time, a conventional single stage actuator will
severely limit the performance of the servo system due to its large inertia and
therefore low bandwidth.
One realistic and emerging solution is dual actuator system. Because the
secondary actuator or micro-actuator can be made small, it has enough bandwidth
to achieve the performance needed for high TPI systems. The question on
micro-actuators in HDD industry is not "if" but rather
"when" it will be used in a HDD. However, the micro-actuator has many
unanswered questions and at present the industry invests much effort into
research of practical implementations.
The purpose of this research project is to focus on the issues
associated with the dual stage servo design. We have begun preliminary studies
on the feasibility of the idea and made some simple implementations on a
dual-stage actuator developed by DSI. Further issues to be studied are the
effect of the limited range of motion of the micro-actuator on the overall
system, the interaction of the two actuators, relative displacement estimator design
and the possible improvement in seek time.
Funding Source: University Academic Research, National University of Singapore, 2000–2003, S$312,175 (with T H Lee
and G Guo)
Remote control and reconfiguration of instrumentation via Internet are
becoming an increasing common event in the workplace. The growth of this
paradigm will parallel the increasing use of telecomputing and telecontrol. The
aim of this project is to build a web-based virtual lab that anyone can access
in anywhere at any time via Internet. Various experiments for engineering
students will be made available in the Web-Based Virtual Lab. With this
advanced facility, students can conduct actual experiments from remote
computers at any time, as if they were working in the actual laboratories.
Researchers in different countries can also conduct research cooperatively and
remotely at the same time.
The project is divided into two phases. Phase one: The methodology of
integration of physical plants, laboratory instruments and the Internet will be
generated. Interface software based on TCP/IP/UDP will be developed to exchange
data between physical plant and web server of web-based virtual lab. Phase two:
on the basis of the first phase, the mechanism to integrate all functional modules
into the web-based virtual lab will be generated. Physical plants and
laboratory instruments are to be connected to web-server. Various experiments
can be conducted remotely.
Funding Source: University Academic Research, National University of Singapore, 1998–2000, S$174,650 (with C C Ko ~ PI)
This project is aimed at developing a Dual Actuator Control strategy
that can be applied to for R/W Head actuation in Magnetic & Optical Disk
Drives. The trend in data storage products are towards high capacity, fast
access, and portability. These trends result in closer track spacing, higher
servo bandwidths, and increased robustness and noise immunity. The reason for a
dual actuator is to use the course stage for fast access and the fine stage for
accurate tracking. CD-ROMs already employ dual actuator for the actuation of
the R/W heads. Hard disks are on the threshold of utilizing dual actuators. The
only reason that has prevented them from doing so is the unavailability of a
cost effective dual actuator.
We believe that this is the opportune time to take a close look at the
dual actuator technology, especially in terms of how to specify the mechanical
plant for a dual actuator for R/W head actuation in storage devices, techniques
for designing its servo system, if possible fabricating one, and demonstrating
that the simulations and the actual performance are in close match.
Funding Source: University Academic Research, National University of Singapore, 1996–1999, S$178,000 (with S Weerasooriya and T H Lee)
In this project we aim to investigate a gain scheduling scheme for an
F-16 flight control system design, in which the nonlinear flight dynamics of
the aircraft is to be linearized at 15 crucial flight conditions. Some newly
developed robust control theory, such as the H2 and H∞ optimization based loop transfer recovery (LQG/LTR), is applied to the
linearized systems to obtain a set of robust observer-based controllers. A gain
scheduling control scheme is then to be proposed to integrate all the designed
control laws together.
Funding Source: DSO National Laboratories, Singapore, 1994–1995, S$5,000 (with T H Lee and E K Poh)