Ph.D. Defense: Peng Wu

Title: Composite Resource Management in Networked Control Systems

Student: Peng Wu

Major Advisor:  Dr. Song Han

Associate Advisors:  Dr. Sanguthevar Rajasekaran, Dr. Bing Wang

Date/Time: Tuesday, December 14th, 2021, 1:00 pm

Location: 

Meeting link: https://uconn-cmr.webex.com/uconn-cmr/j.php?MTID=m0a80fab2963bbbfb36dc9c31ad27e96b

Meeting number: 2621 867 5815

Password: sUJJpJaQ828

Join by phone: +1-415-655-0002 US Toll

Access code: 2621 867 5815

Abstract:

Networked control systems (NCSs) are fundamental to many safety- and mission-critical applications in a broad range of fields such as robotics, process and factory automation. Wide deployment of NCSs requires systematic design tools to enable efficient resource usage while achieving desired Quality of Service (QoS), which might be greatly affected by the inherent imperfections and limitations of the traditional periodic control paradigm, the existing scheduling models, and online resource reconfiguration framework. To address these issues, we explore the composite resource management methods for different scales of NCSs.

For a single NCS, we propose a self-triggered control method based on nonsingular terminal sliding mode control to guarantee the required control quality while minimizing the system resource utilization. To further improve the control performance, we introduce a nonlinear disturbance observer to estimate the external disturbance. With the proof of the system stability, the self-triggering condition based on the control tasks’ inter-execution time is established. This not only enhances the disturbance rejection ability but also allows the control system to sample in lower frequencies with a performance guarantee.

To deal with small-scale NCSs, we introduce a novel model for composite resource scheduling (CRS) in NCSs, which considers a strict execution order of sensing, computing, and actuating segments based on the control loop of the target NCS. We prove that the general CRS problem is NP-hard and study two special cases that are commonly presented in practical settings. The first case restricts the computing and actuating segments to have unit execution time while the second case assumes that both sensing and actuating segments have unit execution time. We propose an optimal algorithm to solve the first case by checking the intervals with 100% network resource utilization and modify the deadlines of the tasks within those intervals to prune the search. For the second case, we propose an exponential-time optimal algorithm based on a novel backtracking strategy to check the time intervals with the network resource utilization larger than 100% and modify the timing parameters of tasks based on these intervals. For the general case, we design a heuristic strategy to modify the timing parameters of both network segments and computing segments within the time intervals that have either network or computing resource utilization larger than 100%.

Along with the ever-growing number of NCSs deployed in the same network and computing infrastructure but managed independently, the complete resource requirements from these NCSs are not known until runtime and may also change during the system operation. There thus does not exist a global scheduler that has the full knowledge of how the resources are allocated among the tasks from individual NCSs. Based on the Regularity-based Resource Partition (RRP) model in a non-uniform multi-resource environment, we study the dynamic acyclic regular composite resource partition reconfiguration (Dynamic ARCRP reconfiguration) problem. We propose the Dynamic ARCRP reconfiguration algorithm with three stages. The initialization stage computes the initial state of the partition system and updates the deadline for each partition. The transition stage then computes the transition schedule based on a heuristic using deferrable scheduling earliest deadline first algorithm. The cyclic stage computes the cyclic schedule based on the constraints from the transition stage.