Visual localization refers to the recovery of theposition and orientation of cameras in known orunknown scenes, using only visual data. We humansuse vision as our primary source of informationfor localization, navigation, and exploration in ourenvironments. Similarly, a high-quality image with awide view of the surroundings often capturessufficient information to represent a locationuniquely. Therefore, using images to localize agentsin a map is an important research area of computervision and robotics. lmmediate consequences ofsolving the visual localization problem leads toexciting and potentially revolutionarymanyapplicationsincluding; humanassistance,autonomoussystems,andaugmented reality.Intuitively, by answering questions such as "Wheream lin the scene?”or “ How to render objects foraugmented reality? ”In this talk, I will present theproblem of visual localization in large scale scenesand recent progress in this direction. The talkcontentswill include,scenereconstruction.understanding,abstraction,andalgebraicframeworks to exploit them forlocalization.Different applications of the visual localizationsystems will also be demonstrate

The problem of verifying whether a high levespecification can be decomposed into localspecifications is of fundamental importancein top-down distributed supervisory controlbut unfortunately the state-of-art verificationalgorithm is of exponential time complexityIn this seminar, we talk about an approachfor reducing its verification complexity andsupporting its parallel verification by usingdivide-and-conquer reductions. We addressthe problemof generatingoptimalreductions automatically. A (partial) solutionto this problem is provided, which consists of1) an incremental algorithm for theproduction of candidate reductions and 2) areductionvalidationprocedure.Astrengthenedsubstitution-basedprooftechnique is used for automatic reductiona fixedvalidation,whiletemplate ofexamples is used forcandidate counterreduction refutation.

In the context of robotics and autonomous systems, theiterative linear quadratic regulator (iLQR) is known to bean effective approach to deal with the nonlineardynamical models in motion planning problems.However, it is the major shortcoming of the iLQR that themethod is not efficient when other complex constraintsare involved, such as the collision avoidance constraintTo overcome this rather significant impediment, a fastand efficient optimization algorithm is developed basedon the alternating direction method of multipliers(ADMM), such that the optimization problem is separatedinto severalmanageable sub-problems and theoptimization process can be accelerated to realizereal-timecomputation. In addition, leveraging thearchitecture of iLOR and neural network, another motionplanning algorithm is developed that avoids the priorknowledge of the system model. Depending solely onmeasurement data, it yields a completely non-parametricroutine for the establishment of the optimal policy. Asclearly indicated from the results attained in severalillustrative examples, these significant merits of theproposed algorithms are demonstrated rather evidently.

Visualizing the flow physics and structural dynamics
is critical to our comprehension of fluid-structure
interaction phenomena, which are pervasive in
engineering and biological systems. Traditional
measurement techniques relyon single-input
sensors and optical diagnostic tools, which often fail
to capture fluid-structure interactions that are
transient in nature, hindering our understanding
of their physical underpinnings. In this talk, i will
present my recent effort in the manufacturing of
multi-input smart sensors that are capable of
capturing velocity and pressure of an airflow, and
quantifying the deformation and UV damage of a
structure. This effort is complemented by the
development of a multi-input optical measurement
technique, which can accurately capture highly
transient fluid-structure interactions. Finally, I will
briefly discuss other techniques I have employed
for the study of fluid-structure interactions, ranging
from evolutionary algorithm-based optimization of
highway infrastructures, to
continuum
mechanics-based modeling of soft active sensors
and actuators, and data-driven modeling of animal
swimming.

The COVID-19 outbreak exposed the vulnerability ofthe global supply chain to a wide audience. Fromtransport and logistics perspective, port congestionshave seriously disrupted the cargo flows in the globalsupply chains. This talk will focus on containershipping industry. Firstly, l will present a few studiesthat we conducted to enhance operation efficiency atport yard and landside that help reduce portcongestion and emissions. This includes optimizingoperations at seaport rail terminals, optimizingcontainer retrieval and reshuffling, and smart stackingat container yard. Secondly, taking the supply chainviewpoint, I will explain the challenging issues incontainer shipping supply chain (CSSC) by focusing ondigitalization and decarbonization. Possible pathwaytoward digitalization andpathwaytowardsdecarbonization in CSSC will be discussed includingfuture research opportunities.

The free-form surface, also termed manifold, is a
smooth 3D skin that contains aesthetic value and
presents functionalities. With the integration of
mathematics, computer-aided design, and
construction techniques, people can bring the
free-form surface into reality. However, how to
achieve an optimal layout on the free-form surface is
still an open topic. In this talk, I will first introduce
my research efforts on topology optimization on the
freeform surface using the extended level set method
and conformal mapping theory. Next, I will present
my work on optimizing a typical adaptive surface- the
origami structure, which can transform from
flattening 2D status to a complex 3D design at a given
crease pattern. In terms of fabricating thin shell
surfaces, I will briefly introduce our surface weaving
method, which can directly construct a 3D surface by
weaving two groups of computed strips. The future
research directions and applications will be discussed
at the end.

The diffraction limit is a fundamental barrier inoptical science and engineering. lt restricts theimaging resolution for both fundamental researchand applications in biology, chemistry, and materialscience industry. Microspheres have beendemonstrated as a powerful platform to challengethe diffraction limit. Microspheres can manipulatelight in a novel way that conventional opticalcomponents cannot achieve. In this presentation.Dr. Chen summarizes his contribution to thefundamental physics,system development,andpotential applications of this technology. The ~20nm observation power was demonstrated forlabel-free samples, which forms the foundation ofopticalsuper-resolutionimaging to observenano-structures in water, oil, and ambient air.

The mission of my research is to answer the question: howcan we provide people with cyber-physical systems (CPSs)they can bet their lives on? Many CPSs are safety-criticalsuch as self-driving cars and critical infrastructures. In thistalk, I will present how my research tackles this problemtowards the goal to enhance the security and safety ofsafety critical CPSs whilst deployed in dynamic, uncertain,and adversarial environments. I will discuss two applicationcases: one is safety assured planning and control for selfdriving cars, the other one is intelligent intrusion detectionfor a water treatment system. For the first application caseI rigorously verify the safety properties of learning-enabledcomponents in autonomous systems and combine formalmethods and optimal control to design safety-guaranteedplanning and control. For the second application case, ldeveloped interpretable symbolic machine learning calledautomata learning to learn hybrid dynamics from dataduring my Ph.D. study. The learned model for a real watertreatment system and intrusion detection results areunderstandable and verifiable for system operators.

Automatically understanding the body posefrom camera inputs promotes many real-lifeapplicationssuchactivityashumanrecognition, autonomous driving,assistantrobotics and sport analysis. This highlydemanding task has seen extraordinaryprogress over the recent years. The success canbe credited to two main factors: effectiveappearance modeling by deep neural networksand the accessibility of large-scale annotateddatasets. However, the current systems are notflawless that still many challenging issues areleft to be alleviated especially when peopleare in complex articulations or several instancesstay close, occluding each other. We argue thatincorporating prior knowledge like the inherentstructure of our body into the network design isequally essential. To this end, in this work, westudy how to design efficient algorithms tojointly optimize the parameters of deep featureextractors and also the probabilistic inferencemodels which encode priors.

介绍轨道交通、道路交通、水运交通以及航空交通等行业的科技发展现状、成就和短板，对国内外交通科技的需求态势和发展趋势进行分析，在总结目前创新发展的基础上，提出交通科技未来发展的目标和总体思路以及各种交通方式的科技发展需求;探讨十四五期间我国交通领域科技创新的主要任务。