This Cyber Physical Systems (CPS) project will develop a theoretical framework that facilitates safe cooperative driving in heterogeneous traffic of human-operated and autonomously-operated vehicles and demonstrate its feasibility through both simulation and physical experiments. This project will help improve the safety of a transportation system currently being transformed by vehicles with growing autonomous features. By introducing an add-on device, or copilot, into legacy human-driven vehicles, this project will offer a smart driving assistant that is aware of the driver's behaviors and can alert the driver when the vehicle is at risk. When engaged in cooperative driving, the copilot will provide advice that reduces the chance of collision with nearby vehicles. By facilitating cooperative driving for both emerging autonomous vehicles and legacy human-driven vehicles, this project will foster a positive attitude of the public toward autonomous driving, therefore accelerating the adoption of autonomous vehicles into the transportation system. The education and outreach activities will raise more awareness of autonomous driving, Artificial Intelligence (AI) and robotics to the younger generation, and stimulate prospective students to pursue degrees and careers in science and engineering.

Human-cyber-physical systems (h-CPS) are interactive engineered systems that collaborate or interact with one or more human beings to leverage the complementary strengths of both human and autonomy technologies. Medical devices, robot assistive systems, teleoperation, semi-autonomous systems, and other technology-assisted applications are all examples of h-CPS. Because human operations are deeply intertwined with cyber and physical processes in h-CPS, new technical challenges for h-CPS analysis and design emerge, particularly in modeling complex human behaviors, enabling effective human-machine interactions, and developing reliable and high-performance controllers. Furthermore, as envisioned in future h-CPS subject to a large amount of data of adequate quality and quantity available from rich sensing modalities, modeling, interaction, and control procedures are shifting from model-based to data-driven, and new challenges such as trustworthiness and learning efficiency of data-driven methods are expected to arise. This project targets the unique challenges of data-driven modeling and control of h-CPS by developing a holistic data-driven design framework, accounting for addressing today's major barriers to apply data-driven approaches for modeling, interaction, and control of h-CPS. Educational and outreach activities are well-integrated into the research and include CPS workforce training, interdisciplinary research and curriculum development, and K-12 STEM outreach activities. The designed activities are uniquely positioned to attract members of underrepresented groups with a focus to enhance the diversity of the federal, state, and local CPS workforce.

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).

Cyber-Physical Systems (CPS) sustainably benefit from software upgrades throughout their life cycles. However, as CPS become machine-learning-intensive due to rapidly increasing interactions between CPS and machine learning technologies, two major distinguishing factors associated with machine learning techniques raise significant safety concerns about CPS upgrades which play a critical role in enabling lifetime safety assurance. First, upgrades of machine learning components, which inherently result in system changes, come at significant safety risk for safety-critical CPS due to the vulnerabilities of machine learning techniques. Second, the traditional safe-by-verification upgrade framework, in which upgrades and verification have to be two separate procedures, is no longer valid for machine learning processes that update instantaneously during system operations. This project targets these unique challenges by developing scalable verification and monitoring methods for upgrades as well as safe upgrade procedures to enable trustworthy upgrades and achieve lifetime safety assurance in machine-learning-intensive CPS.

Deep Neural Networks (DNN) enabled Cyber-Physical Systems (CPS) hold great promise for revolutionizing many industries, such as drones and self-driving cars. However, the current generation of DNN cannot provide analyzable behaviors and verifiable properties that are necessary for safety assurance. This critical flaw in purely data-driven DNN sometimes leads to catastrophic consequences, such as vehicle crashes linked to self-driving and driver-assistance technologies. On the other hand, physics-model-based engineering methods provide analyzable behaviors and verifiable properties, but do not match the performance of DNN systems. These considerations motivate the work in this project which aims at physics-model-based neural networks (NN) redesign, yielding HyPhy-DNN: a hybrid self-correcting physics-enhanced DNN framework. HyPhy-DNN will provide the performance of DNNs and the analyzability and verifiability of physical models, thus providing a foundation for verifiably safe self-driving cars, drones, and other CPS systems. Moreover, the HyPhy-DNN will fundamentally advance the integration of deep learning and robust control to enable safety- and time-critical CPS to safely operate with high performance in unforeseen and dynamic environments.

Human-driven vehicles (HDVs) and automated vehicles (AVs) of all levels (Level 1-5, AVs1-5) may share the highways in the long and foreseeable future. The increasing vehicle autonomy heterogeneity and diversity may jeopardize the safe and harmonious interaction among such vehicles with mixed autonomy on highways and pose a threat to the safety of all vehicles. This may exacerbate an already growing and alarming national concern on traffic safety. This project aims to advance the state of the art in the Cyber-Physical Systems (CPS) research areas of Autonomy, Safety, and Transportation by ushering in a new CPS paradigm of harmonious and safe integration of highway vehicles with heterogeneous, varying, and mixed human / machine autonomy. Through collaborative research, the project may create new methods and tools to enhance the highway driving safety of heterogeneous vehicles. The outcomes of this work may also be extended to advance other CPS in manufacturing, warehousing, and healthcare applications where interaction among humans and heterogeneous autonomous robots is pervasive and safe coordination among them is critical.

Autonomous vehicles (AVs) are revolutionizing the transportation ecosystem and are expected to become a critical part of our society. AVs are equipped with many electronic devices, including various sensors, electronic control units (ECUs), internal control networks, as well as capabilities in artificial intelligence, computing, storage, and communication. Although the automotive industry, as well as the public, are optimistic that an AV can perform many basic functions on par with human drivers, few are confident about the security and safety of AVs, especially when AVs are highly vulnerable to potential attacks from cyberspace, as demonstrated in recent series of car hacking incidents. In this project, a team of researchers from Virginia Tech aims to address some of the fundamental security and safety challenges for AVs. The research team follows a novel defense-in-depth approach that combines three layers of defense against attacks on software systems, in-vehicle networks, and safety-critical ECUs in an AV. Each layer can be designed and deployed independently from the other layers and when working jointly, they can not only effectively thwart most system and network attacks but also provide fail-operational protection against both known and potentially unforeseen cyberattacks.

Autonomous driving is on the verge of revolutionizing the transportation system and significantly improving the well-being of people. An autonomous vehicle relies on multiple sensors and AI algorithms to facilitate sensing and perception for navigating the world. As the automotive industry primarily focuses on increasing autonomy levels and enhancing perception performance in mainly benign environments, the security and safety of perception technologies against physical attacks have yet to be thoroughly investigated. Specifically, adversaries creating physical-world perceptual illusions may pose a significant threat to the sensing and learning systems of autonomous vehicles, potentially undermining trust in these systems. This research project aims to deepen our understanding of the security and safety risks under physical attacks. The project endeavors to bolster sensing and learning resilience in autonomous driving against malicious perceptual illusion attacks. The success of the project will significantly advance the security and safety of autonomous driving in the face of emerging physical-world threats, paving the way for the safe deployment of autonomous vehicles in next-generation transportation systems.

In the current state-of-the-art machine learning based real-time control of large complex networks such as electric power systems is largely bottlenecked by the curse of dimensionality. Even the simplest control designs demand numerical complexity to accomplish. The problem becomes even more challenging when the network model is unknown, due to which an additional learning time needs to be accommodated. This project will take a new stance for solving this problem, and develop a suite of hierarchical or nested machine learning-based schemes that take advantage of various forms of physical redundancies in the network dynamics to learn only the most important traits of its behavior instead of wasting time in learning minor traits that may improve the closed-loop performance only by a small amount. This selective learning approach will reduce learning time by several orders of magnitude, making real-time control more tractable and more implementable. Products will include numerical algorithms that are applicable across a wide range of machine learning based control. In terms of societal impact, the project is strongly envisioned to bring control theorists closer to data scientists so that these two research communities can work together, and answer important questions such as: why the value of big data has traditionally been under-utilized in controls, what new dimensions can control theory gain from machine learning and vice versa, and what primary analytical and experimental tools are needed to make this marriage more successful. The research will also support the cross-disciplinary development of a diverse cohort of PhD and undergraduate students, and the development of a graduate-level course on the applications of machine learning in control.

This project investigates new reinforcement learning (RL) approaches for cyber-physical autonomy to bridge the gap between current intelligent systems and human-level intelligence. The nature of many cyber-physical systems (CPS) is distributed, heterogeneous, and high-dimensional, making the hand-coded functions and task-specific information hard to design in the learning scheme. Large amount of training data is often required for achieving the desired performance, however this limits the generalization to other tasks. Hence, this project is to explore the new RL strategies to enable CPS with the capabilities of autonomous learning and generalization to rapidly adapt in unknown situations that were not assumed in the design phase. The results are expected to transform how agents interact in high-dimensional and heterogeneous environment, and therefore could potentially provide in-depth findings for exploring creativity in frontier Artificial Intelligence techniques.

Unmanned Aircraft Systems (UASs), or drones, have tremendous scientific, military, and civilian potential for data collection, monitoring, and interacting with the environment. These activities require high levels of reasoning, perception, and control, and the flexibility to adapt to changing environments. However, like other automated agents, UAS don't possess the ability to refocus their attention or reallocate resources to adapt to new scenarios and adjust performance. This project will provide a new class of control and planning algorithms capable of adjusting performance as computing resources are continually reallocated, such as when transitioning from waypoint navigation to environmental sample collection. A computing framework to make use of freed resources will be developed allowing autonomous agents to focus attention where it is needed, for example, away from navigation and to perception. Together, these will provide a blueprint for making use of similar algorithms with adjustable performance (e.g., anytime algorithms) which can be adapted to other robotics platforms, as well as water, space, or ground vehicles.

These technology innovations will improve the ability of agents to learn more, perceive more accurately, collect better data, and respond more appropriately to changing environments and mission objectives. Specific to UAS, this project will help maintain U.S. air superiority goals through agile planning, targeted and persistent Intelligence, Surveillance, and Reconnaissance (ISR), and flexibility and adaptability. The project goals are coupled with outreach and educational activities focused on increasing the understanding of rural populations of the value of investing in scientific and technological research. The educational efforts, targeted at K-12, undergraduate, graduate, and adult engagement are designed to dramatically increase the CPS educational pipeline in the Midwest.


The project focuses on achieving its goals by providing a complete framework for a class of performance-adjustable, resource-aware algorithms called "co-regulation." First, a new modeling and analysis framework, Co-regulated Hybrid Systems (CHS), will provide a mathematical foundation for optimal control, control synthesis, and performance analysis for systems that can dynamically vary sampling rate and other computational resources to adjust performance. Next, using the CHS formalism, computational workload is predicted forming the basis for a novel Co-regulated Real-Time Kernel (CRTK) to dynamically reallocate computing resources while guaranteeing real-time schedule feasibility. Finally, a co-regulated Markov Decision Process (MDP) forms the planning portion of a resource-aware autopilot for adaptable UAS. The system will be implemented in a multi-agent, rainforest monitoring scenario requiring periods of surveillance, sampling of plants, and emplacement of sensors.