Come join this Information Webinar to learn more about:the benefits youwill gain from studying research postgraduate programs;why choose HKUST to pursue your MPhil or PhD studies;how HKUST research postgraduate programs will put you onto a promising path;the latest development of HKUST(Guangzhou) and admission to Guangzhou Pilot Scheme; and the first-hand experience from our International research postgraduate students.

China Intelligent Manufacturing 2025 is a vital national strategy for developing China as a world super power in manufacturing, which can make everything with high quality and efficiency. In this strategy, the critical but most difficult objective is to build smart machine tools. Smart machine tools can efficiently cut workpieces into qualified parts without manual operation. On these machines, cutters can be automatically measured for wear with a sensor (an on-machine tool setter) during machining, and then the following toolpath is compensated with the tool wear. Besides, the workpiece is measured with a sensor (a touch probe) right after each geometric feature is cut, and re-machining of the out-of-tolerance area is automatically planned. However, smart machine tools have not been built and applied in the manufacturing industry. It is urgent and important for Chinese researchers to develop the kernel techniques of smart machine tools. Prof. Chen is leading an international research team to conduct advanced research on smart machine tools in the following topics: (a) automatic multi-axis tool path generation and compensation, (b) automatic, in-process and on-machine cutter inspection, and (c) automatic, in-process and on-machine workpiece measurement. His team has successfully addressed many technical challenges and developed several new products for smart machine tools.

We learn and remember multiple new experiencesthroughout the day. The neural principles enablingcontinuous rapid learning and formation of distinctrepresentationsofnumeroussequentiaexperiences without major interference are notunderstood. To understand this process, weinterrogated ensembles of hippocampal place cellsas rats explored 15 novel linear environmentsinterleaved with sleep sessions over continuous16-hour periods. Remarkably, we found that apopulation of place cells were selective toenvironment orientation and topology.Thisorientation selectivityproperty biasedthenetwork-level discrimination and re/mappingbetweenmultipleenvironments.Novelenvironmental representations emerged rapidly asmore generic, but repeated experience within theenvironments subsequently enhanced theirdiscriminability. Generalization of prior experiencewithdifferentenvironmentsconsequentlyimproved network predictability of future novelenvironmental representations via strengthenedgenerative predictive codes. These coding schemesreveal a high-capacity, high-efficiency neuronalframework for rapid representation of numeroussequentialwithexperiencesoptimaldiscrimination-generalization balance and reducedinterference.

Robots are being increasingly deployed in the real world for applications such as warehouse automation, search and rescue, surveillance and reconnaissance, and environmental monitoring. However, when performing long-term tasks, robots need to deal with uncertain, changing environmental conditions. Robots may fail or if they are operating in adversarial environments, they may get attacked.I will describe resilient coordination algorithms to cope with adversarial attacks and applications to information gathering and target tracking with aerial robots. In addition, I will show how to use ideas from stock portfolio optimization to build risk-aware robot teams that can balance the trade-off between risk and reward. Further, I will describe how to employ graph neural networks to enable large-scale, decentralized multi-robot coordination and planning. In the end, I will discuss some future work on securing robot teams against adversarial attacks and uncertainties when the robots use machine learning techniques for perceptions and communications.

Car-following is the mostcommontask. lt refers to a process where the followingvehicle (FV) tries to keep a safe distance betweenitself and the lead vehicle (LV) by adjusting itsacceleration in response to the actions of thevehicle ahead. The corresponding car-followingmodels are functions that determine FV's futureaccelerationsbased on current (and historicaldriving situations. Car-following models are thecornerstone for microscopic traffic simulation andintelligent vehicle development. One majormotivation of car-following models is to replicatehuman drivers'longitudinal driving trajectories. Inthis seminar, will first talk about how to calibrate
classical
evaluate,andcross-comparecar-following models using large-scale real-worldnaturalistic driving data. To model the long-termdependency of future actions on historical drivingsituations, will also introduce a long-sequencecar-following trajectory prediction model based onthe attention-based Transformer model.

With over half a billion years of evolutionary history, fishes swim with high efficiency, great agility, and surprising stealth in the three-dimensional aquatic environment. Therefore, it is undoubtedly natural for engineers to turn to fish as a source of new ideas on underwater propulsive systems.

Over the past decades, roboticists, inspired by these biological systems, have constructed various fish-like robots that copy real fish morphology, locomotion, and movement.

Recently, the direction of bio-inspiration has started to flip to bio-understanding----using robotics in engineering to help answer questions in biology.

In this talk, I will first introduce how we construct and control the robotic fish following the paradigm of bio-inspiration. Then, I will give examples of applying real and virtual robots to help us explore why and how do fish school. Finally, I will briefly introduce my ongoing and future work in bio-inspiration and bio-understanding.

Rail is safe, efficient, and environmentally friendly formoving large volumes of passengers and freight overlong distances. Safety is the top priority for global railsystems. In the age of Industry 4.0 empowered by theuse of the Internet of Things (loT), advancedcommunications technologies, artificial intelligence(Al),andautomation technologies, the railtransportation system is stepping onto a new phase ofrapidproactive,digital,developmenttowardintelligent safety and risk management. Thispresentation will introduce several cross-disciplinary.rail-centric A and data analytics-related studiesconducted at the Rail and Transit Program at RutgersUniversity, in close collaboration with the U.S. railindustry. Particular examples include predictivemodeling of rail infrastructure degradation, as well asreducing fatalities and injuries on railways through theuse of Al and communications-based advanced traincontrol technologies.

The performance of semiconductor integrated
circuits can be improved through 3Dintegration
of additional layers in the post-Moore era.
However，3Dintegrationofoptoelectronics
remains a technological challenge.3Dprinting
has been emerging as a disruptive approach to
manufacturing due toitssimpledesign-friendly
manner. Over the past decadesscientistsand
engineers have continually developed methods
for a diverse selection of materialshigh
precision and throughput that are essential for
3D optoelectronic devices but so far this still
remains a great challenge. In this talkIwill
introduce my
study
of
developing
high-performance 3D nanoprintingtechniques
for advanced 3Doptoelectronics.Thestudy
consists of two parts:(1)developmentof
advanced 3D printing techniques for functional
semiconductor materialswhere nanopipette
confined supersaturation is localized for creating
freeform perovskites 3D nanostructures as3D
nanopixels
and
heterostructures.(2)
Development of parallel 3D nanoprinting in a
liquid meniscus or a droplet form todeliver
materials ink that improves the throughputThis
study will enable a true manufacturing platform
for next-generation 3D optoelectronic devices.
making a great impact oninterdisciplinary
manufacturing and materials science fields such
as stretchable electronicsphotonicsand
biomedical engineering.

Neuromuscular junctions (NMJs) are synapses formedbetween motoneurons and muscle fibers, andcharacterized with high density of acetylcholinereceptors (AChRs) at postsynaptic membrane. Rapsyna postsynaptic adaptor protein, anchors AChR to thecytoskeleton and induces AChR cluster formation. Ourrecent result shows that rapsyn is an E3 ligase whichneddylates subunits of AChR and promotes AChRclustering. However, how these processes are regulatedremains unclear. To this end, we firstly focus on rapsynmutationsidentifiedin congenital myastheniasyndromes (CMSs)，a group of heterogeneous NMJdisorders characterized with muscle weakness andfatigues. Through studying disease-related mutation.we found that rapsyn become tyrosine-phosphorylatedin response to upstream signaling. Phosphorylation ofrapsyn promotes its self-association and E3 ligaseactivity and thus promotes AChR clustering, suggestinga signaling molecule of rapsyn. Moreover, we combinedbiochemical assay with cellular culture and in vivoassay and found that rapsyn undergoes liquid-liquidphase separation (LLPS) and condensates intoliquid-like assemblies. Such assemblies can recruitAChRs, cytoskeletal proteins, and signaling proteins forpostsynapticdifferentiation, revealing a novelmechanism of rapsyn in postsynaptic assembly.

A mircoLED display is a self-emissive display
comprising millions of micrometer-scale LEDsIt
potentially outperform LCD and OLED display not
only because its higher power efficiency, better
image quality, wider color gamut but also due to its
good scalability, transparency, flexibility and other
characteristics. lt is an emerging technology
expected to disrupt the display market. For
manufacturing microLED displays, tens of millions
of microscopic LED chips need to be transferred
and assembled at a high through-put with an
ultra-high yield. There is significant investment
across both industry and academia, and the
technology is advancing very quickly in this field;
however, there is no existing approach that has
demonstrated a cost-effective way of doing it. in
this talk, I will introduce the background of the
application and review state-of-the-art mass
transfer development for microLED display
manufacturing. Methods including elastomeric
stamping, fluidic assembly, laser printing, etc. will
be discussed and assessed in details.