Networked systems, e.g., communication and logistic networks, or distributed computing and control environments, incorporate separate yet cooperating entities to achieve the intended goals. The inevitable uncertainties related to the information and resource exchange among the system components – congestion, noise, channel interference, node and interface failures – may compromise the desired functionality. Virtually all the essential parameters of the transfer process – delay, error rate, throughput – change in a dynamic and unpredictable way. In order to leverage the impact of network-induced and environmental perturbations, one may simultaneously engage multiple transport channels. Unfortunately, inappropriate traffic distribution in the multi-flow framework may actually cancel the advantages of increased efficiency and resilience this option offers.The project objective is to:
1) identify the common characteristics of modern networked systems that explicitly involve multiple channels for resource and information exchange and, using advanced methods of control theory,
2) develop new algorithms governing the multi-flow transport process that will yield better robustness and higher efficiency than the existing solutions according to formally defined quantitative criteria.
The proposed approach emphasizing the system dynamics, besides improving the operation of a given class of systems, is believed to relieve the transition of the outstanding results to other application areas and problem types, e.g., adopting data flow control mechanisms from computer networks to resource distribution in logistic systems, in line with the concepts of Physical Internet and Industry 4.0. The designed new algorithms are to provide a high degree of resiliency with good dynamics while conforming to the class-specific constraints, network topology, and information exchange hazards. Therefore, contrary to the solutions delivered in the past, the optimality criteria will be given equal attention as the implementation and robustness aspects in the design process. In order to corroborate the theoretical outcomes, a test platform for the control system analysis and convenient yet rigorous algorithm property evaluation in relation to the current solutions will be created. Unlike the previously reported endeavors in the field, this platform should allow for conducting comprehensive studies at all stages of the design process – from concept validation, through parameter selection, to experimental verification – in the optimal (efficiency) and robust (sensitivity) contexts. The created test environment will encompass both new modules and programs developed for the professional simulation tools, like Matlab or ns-3, and experimental setups with hardware units and benchmark plants cooperating via real data transmission networks.