University of Liverpool -- National Tsing Hua Unviersity Dual PhD Project
The position can be started as soon as possible. We will consider applications until the position is filled.
A fully funded 4-year PhD position is available as a joint program between the Department of Computer Science of the University of Liverpool (UoL), UK, and the Department of Computer Science of the National Tsing Hua University (NTHU) at Taiwan, ROC.
The CS department at UoL is a world-leading research centre with strengths in robotics and intelligent systems,
automated verification, algorithms and optimization, and computational economics.
The CS department at NTHU is internationally known for its excellence in research and teaching,
with its research focusing on computer and communication research, IC design technology,
and image processing and computer vision. The successful candidate will conduct research
in the areas of automated verification, computer vision, and robotics.
He/she will be jointly supervised by Dr. Xiaowei Huang (https://cgi.csc.liv.ac.uk/~xiaowei/) at UoL and
Prof. Shang-Hong Lai (www.cs.nthu.edu.tw/~lai/) at NTHU.
The PhD project will focus on theoretical foundation and algorithmic development of verification
techniques for the machine learning models used in robotic perception and manipulation.
The ability to recognize and manipulate objects is central to robotics. When interacting
with human users, a robot may be asked to recognize a certain object and determine if
and how to safely grasp the object in order to fetch it. With the availability of large
scale training datasets, deep learning models have been trained for this task and may
be deployed in safety critical applications. However, when operating in noisy (benign) or
adversarial (malicious) environment, the sensor reading of the robot may be subject to small
perturbations. It has been known that small, or even imperceptible, perturbations may lead to
failure of object recognition. The first objective of this PhD project is to develop
quantitative evaluation techniques, based on probabilistic verification, to check the
robustness of deep learning based object detection techniques against small perturbations.
Based on this, the second objective is to develop verification techniques to check the safety
of robots' perception-action loop, in which the interactions with the environment and the human operators will be considered.
The candidate will study at NTHU in the 1st and 4th years and at UoL in the 2nd and 3rd years.
During the stay in NTHU, he will become a member of Computer Vison Laboratory. During the stay
in UoL, he will become a member of Robotics and Autonomous Systems Group.
The position is fully-funded. When the student is at Liverpool, the UoL will pay both tuition fees (at appropriate rate for Home/EU or OSI student), and stipend (£256 per month, or £3072 per year).
NTHU will do the similar when the student is in Taiwan.
Please feel free to contact me for any question.