In cellular communications, the mobile phone and the base station need to adapt their model of the channels connecting them in order to remove reflected signals. This adjustment is made with every data packet and the overhead is that about one seventh of all the data traffic is known to both ends and so is information-free. That is the price of adaptation. The benefit is that the call remains clear despite the motion of the cell phone.
In network communications, the environmental variable is the competing traffic and the data send-rate serves two purposes; to convey data to the other end, and to probe the network to estimate the maximal throughput. A significant portion of the capacity for transmission is consumed in assessing these limits.
The objective of this research is to explore the trade-offs present in such adaptive systems so that more highly performing approaches can be developed. How much capacity must be lost to achieve effective use of the variable channel? By commencing with a familiar system, we hope to recognize both the design adjustments already made and the opportunities for practical advances based on new theory.
The intellectual merit of the work lies in extension of technical concepts from systems theory to the realm of random evolution of networks, such as describe the traffic behavior. This brings new perspectives and mathematical tools into that are rooted in engineering practice.
The broader impact of the research is its connection to people?s experience. Today?s children are very familiar with the variability of download speeds and the behavior of computers to adjust to this. Thus an entrée is provided to involve students in hard theory to address everyday problems and therefore engage them early in the value provided by technical sophistication.
Information and information quality, that is the reliability or noisiness of information, are examined in the context of feedback control systems which adapt to their environments. A well-known example of radio power control in mobile wireless is used as a test area. New approaches were developed which rely on extraordinarily high computing power to yield an optimal solution. These then yielded heuristics of new approaches to power control. The energy and building sectors were also studied to develop new approaches to deal with forecast data in feedback control. The property that forecasts diminish in reliability with horizon was incorporated into the formulation to derive new techniques which accommodate the data quality. In these applications sectors, this will permit the posing of the question: "what performance benefit might we see for installing improved sensors to allow better quality forecasts?". These approaches have already drawn some interest from industries in the energy sector to assist in the more effective operation of generating machinery in both the renewable and non-renewable areas. Outreach from the project has been to high-school students and to industry with the focus being on the design aspects of modeling based on data and the need for informative experiments. This preserves the human aspects of system design in understanding and managing data quality, which is not yet an aspect amenable to machines.
