The development of efficient and scalable software systems is crucial for the growth of Industry 4.0 and its related applications, such as smart manufacturing and design technologies. In this talk, I will share our projects on developing such systems for automated robot operation, manipulation, and surveillance, with applications in diverse components of Industry 4.0. First, I will introduce our research framework, which focuses on the localization and perception of robotic systems.

The development of Autonomous vehicles (AVs) has potential benefits by reducing the number of road fatalities and by improving the accessibility of mobility services. However, these benefits can only be achieved when drivers maintain a calibrated level of trust towards AVs. This report is going to: 1) discuss the definitions and models of human-automation trust; 2) introduce the critical roles of trust in AV acceptance and possible antecedents of trust; 3) discuss trust assessment with multimodal physiological signals.

Small-scale robots have great potential to revolutionize modern medicine by enabling non-invasive medical interventions at hard-to-reach and delicate body locations. However, these robots are inherently limited in actuation, computation, control, and power storage due to space restrictions arising from the miniaturization of conventional robotic components, such as actuators, processors, and sensors. Therefore, it is crucial to develop new original design and fabrication strategies different from those used to build traditional large-scale robots to advance small-scale robots and facilitate their translation from benchtops to bedsides.

Driven by the development of Internet of Vehicles (IoV) and artificial intelligence, enormous smart vehicles are ubiquitously connected to the Internet through cellular networks, with the potential of supporting a plethora of computation-intensive and delay-sensitive transportation applications, such as object classification and autonomous driving. To enhance the computing capability of hardware-constrained vehicles, mobile edge computing (MEC) has recently emerged as a cost-effective paradigm by releasing and distributing computing resources to the edge servers within the radio access networks. In this talk, we focus on pursuing optimal task offloading and resource allocation in MEC by leveraging deep learning and statistical learning techniques.

Road transportation is a complex system with complicated interactions among human, vehicles, and infrastructures. Its emissions, including Greenhouse Gas (GHG) and air pollutants, lead to serious environmental problems that further affect human health and social equity. This seminar will start with a brief introduction on basic terminologies of road transport emissions. With the understanding of the issue, the quantification and characterization methods to analyze vehicle emission and energy consumption will be introduced, and the application scenarios of different model types will be compared. Strategies to mitigate road transport emissions and air quality problems, including vehicle connection & automation, electrification, and travellers’ behavioural guidance, will be proposed and evaluated. As a final remark, policy implications and future research will be discussed.

The central nucleus of amygdala (CeA) plays key roles in a diverse set of adaptive behaviors, including defensive and appetitive response. Early studies of CeA emphasized its function in aversive Pavlovian learning, it was described as a relay connecting basolateral complex of the amygdala to the brainstem effectors. With methodological advances, we demonstrated the somatostatin (SST) expressing neurons in the CeA was actively encoding fear memory through fiber photometry in mice undergoing fear conditioning. Closer observation activities of individual SST+ CeA neurons with endoscopy imaging revealed their roles encoding both aversive and appetitive salient events. In parallel study of PKC-δ -expressing neurons, another major population of the CeA, reveals they are essential for the synaptic plasticity underlying learning in the lateral amygdala, which challenges traditional model of fear learning.

Molecular dynamics (MD) simulations are statistical mechanics-based tools for predicting microscopic and macroscopic properties/dynamics based on the details of molecule-molecule interactions. They have been widely applied in biomedical and engineering fields. The application of MD on human b defensins (hBDs) type 2 (hBD-2) and type 3 (hBD-3) is covered in this talk. hBDs belong to the human innate immune system and have microbicidal activities against bacteria, fungi, yeast, viruses, in addition to having chemotactic activities. In order to understand its functional mechanism in molecular detail, the binding structure of hBD-3 on model bacterial membrane was predicted using microsecond-long MD simulations and validated by solid-state NMR method. It was found that hBD-3 binds with bacterial membrane on the two loop regions. Free energy perturbation (FEP) calculation further clarified the contribution of each residue to the membrane binding, which showed that the tail region of hBD-3 drives the process. The results emphasized the importance of electrostatic interaction in hBD-3’s binding with bacterial membrane. Besides that, it was found that hBD-2 can bind on the receptor binding domain (RBD) of SARS-CoV-2 (Cov-2) at the same site as the receptor angiotensin-converting enzyme 2 (ACE2), thus blocking CoV-2 virus from attacking ACE2 expressing cells. The result supplies insight on the role human innate immune system plays in combating CoV-2 infection.

Nature systems having a body with many degrees of freedom are often considered the ultimate model for machines. To confer the performance advantage of animal systems on robotic machines, we are conducting studies on a thoroughunderstanding of the biological systems at the biomechanical level and the developments of biomimetic intelligent machines and biologically inspired robots. Unfortunately, many of these robots are crucial for investigation but have yet to be widely applied to real applications. We, however, can use the insight from the study of biologically inspired robots to develop robotic machines for possible applications.
In this talk, I would like to introduce some of our studies on biologically inspired robots first and then provide some designs of environment-adaptive robots for real applications.

1. 深圳市地面公交财政补贴政策研究2. 公共交通导向下的建成环境资源配置策略3. 交通运行监测中心与调度平台的北京实践4. 交通拥堵防治决策系统与技术的广西实践

Over the past decade, programmable Cas9 nucleases have revolutionized basic research and clinical applications. However, compared to Cas9-based DNA targeting tools, there is a lack of simple and effective tools for studying and manipulating RNA. The Type VI CRISPR-Cas13 system is currently the only known efficient programmable system exclusively targeting RNA. My research is focused on the type VI-D effector protein Cas13d, which is one of the smallest effectors in the Cas13 family, making it more advantageous for packaging into adeno-associated viruses. We utilized cryo-electron microscopy (cryo-EM), biochemical approaches, and hydrogen-deuterium exchange mass spectrometry to explore the conformational landscape of Cas13d during RNA processing, and utilized the structural insights to guide Cas13d protein engineering. Additionally, we established in vitro and cellular assays of using Cas13d to achieve mRNA knockdown, pre-mRNA splicing regulation, and RNA detection, providing proof-of-concept for future RNA-centric applications. Overall, our findings have the potential to significantly advance the field of transcriptome engineering and lead to more effective gene therapies.