Research

Vehicular Networked Systems (VNS) are Cyber-Physical Systems (CPS) where the communication, control and information processing technologies are integrated to achieve more safe and effective transport for autonomous vehicles. Examples include autonomous underwater vehicle systems, intelligent air transportation systems, smart transportation systems and smart factory automation systems. Resilient VNS has abilities to maintain situational awareness of their surrounding threats and take actions assuring a return to operational normalcy in a safe and efficient way.

Research challenges:

1. wireless networks that promise to provide improving safety are inherently unreliable and subject to deep fading and limited bandwidth. This unreliability will severely limit, degrade the information exchange between the vehicles and cause serious safety issues.

2. temporal variations in the quality of service of these wireless networks are strongly coupled with the dynamics of the controlled physical systems. The changes of many critical control variables in VNS, for instance, the distance, orientation or velocities of the vehicles, will significantly affect channel states, such as data rate, channel coherence time. This impact on communication channels will in turn penalize the controller performance, most likely leading to unsafe actions on vehicles.

3. the open-access nature of wireless networks introduces many vulnerabilities for malicious cyber attacks. To guarantee safe operation, the resilient VNS must be able to agilely response to those attacks and adaptively reconfigure existing strategies.

A human-machine system consists of human operators and automated machines that interact/collaborate with each other to accomplish complex tasks that are difficult for either humans or machines alone. Such human-machine systems, however, are subject to system failures caused by large and unexpected human uncertainties. The figure on the left shows a hybrid assembly example where assembly tasks are allocated to both human operators and automated robots. Prior work showed that human factors, such as human trust and fatigue, have paramount impact on human performance in task execution. Thus, unlike the resilience issues in unmanned autonomous systems, the resilience of a human-machine system relies on the resilience of machine and human as a whole, and therefore demands interdisciplinary approaches from both the engineering and psychology communities.

Industrial processes are typically complex systems with high nonlinearity. The complexity lies in the fact that the process systems often consist of many subsystems having different dynamics that are physically connected. The system dynamics of these processes are difficult to be modeled using priori first-principle due to their high nonlinearity and various operational conditions. On the other hand, with the rapid development of computing, sensing and storing technologies, oceans of recorded process data become available and motivate data driven based approach as an attractive alternative for system monitoring, control and optimization.

Research challenges:

1. The inherent complexity brought by various levels of dynamics in industrial processes systems leads to challenges of modeling the industrial systems using traditional priori first-principle methods. The modeling challenge motivates the development of new data-driven based modeling frameworks that can effectively extract system information from excessive data generated in the process.

2. With the data-driven based framework, the second challenge lies in how to design control strategies under which the complex industrial processes attain quality performance.

3. The profit of industrial processes often depends on the transition behavior among various operation conditions. The data-driven based modeling approach enables the characterization of such transition behavior. The third challenge is about how to regulate the transition behavior to achieve the maximum profit.