The objective of this research is the development of novel control architectures and computationally efficient controller design algorithms for distributed cyber-physical systems with decentralized information infrastructures and limited communication capabilities. Active safety in Intelligent Transportation Systems will be the focus cyber-physical application. For the successful development and deployment of cooperative active safety systems, it is critical to develop theory and techniques to design algorithms with guaranteed safety properties and predictable behavior. The approach is to develop a new methodology for the design of communicating distributed hybrid controllers by integrating in a novel manner discrete-event controller design and hybrid controller design and optimization.
The methodology to be developed will exploit problem decomposition and will have significant technological impact for a large class of cyber-physical systems that share features of modularity in system representation, partial information, and limited communication. The focus on distributed control strategies with limited communication among agents is addressing an important gap in existing control theories for cyber-physical systems. The approach will mitigate the computational limitations of existing approaches to control design for hybrid systems.
Given the focus on cooperative active safety in Intelligent Transportation Systems, the results of this effort will have significant societal impact in terms of increased traffic safety and reduced number and severity of accidents. The broader impacts of this proposal also include involvement of high-school and undergraduate students and curriculum development by incorporating results of research into existing courses on cyber-physical systems.
The principal objective of this project was the development of novel control architectures and computationally efficient controller design algorithms for distributed cyber-physical systems with decentralized information infrastructures and limited communication capabilities. The cyber-physical application that served as a platform for the validation of the results is that of cooperative active safety in Intelligent Transportation Systems. As further advances are made in communication and sensing technologies, Intelligent Transportation Systems are expected to become more comprehensive. We envision a near future in which vehicles will be equipped with on-board active safety systems that predict and prevent vehicle crashes decreasing traffic accidents and fatalities. In this project, we specifically studied collision avoidance at traffic intersections. The work developed under this award established fundamental theory to ensure safety in multi-agent systems under information limitations such as coming from communication delays and noisy sensors, as is the case when multiple vehicles approach a traffic intersection. A significant challenge is to tackle the computation complexity that arises when multiple agents are involved. We identified techniques that lead to algorithms that scale polynomially with the number of agents, allowing the concrete application of these techniques to the design of supervisors for driver assist systems that prevent collision at traffic intersections. The algorithms were implemented in a simulation tool with visualization capabilities (see images). In addition, experimental demonstrations were carried in the Multi-Vehicle Laboratory of the Co-PI at MIT (see image). Intellectual Merit: Novel methodologies for synthesis of safe controllers for a class of distributed cyber-physical systems were developed. They are based on constructing a suitable discrete abstraction of the continuous system model and on solving the controller synthesis problem on the resulting discrete-event abstraction using the theory of supervisory control of discrete-event systems or the theory of scheduling. These methodologies can handle both disturbances and measurement uncertainty in the continuous state equation and they achieve better scalability than prior approaches. Broader Impacts: The award fully supported one graduate student at the University of Michigan (UM), partially supported two more at UM and MIT, and partially supported three post-doctoral researchers at MIT. REU supplements supported a large number of undergraduate students at UM and MIT, many of whom have gone on to graduate school. The project's impact has been enhanced by the laboratory test-bed at MIT. The PIs presented the accomplishments of this project at several institutions in the US, France, and Italy. The methodology developed in this project is being incorporated in graduate courses at UM and MIT.
