EditOverview
Although the Information Technology (IT) transformation of the 20th century appeared revolutionary, a bigger change is on the horizon. The term Cyber-Physical Systems (CPS) has come to describe the research and technological effort that will ultimately allow interlinking of the real-world physical objects and the cyberspace efficiently.
The integration of physical processes and computing is not new. Embedded systems have been in place for a long time and these systems often combine physical processes with computing. The revolution will come from massively networked embedded computing devices, which will allow instrumenting the physical world with pervasive networks of sensor-rich embedded computation.
As Moore's law continues, the cost of a single embedded computer equipped with sensing, processing and communication capability drops toward zero. This makes it economically feasible to deploy networks of such nodes to take a very large number of sensor readings from the physical world and compute quantities and take decisions out of those sensor readings.
A large number of sensor readings is useful because (i) it can cover a large area and (ii) a dense network can offer better resolution and has therefore a better capability to detect the occurrence of an event.
Applications are however typically not interested in all sensor readings; they are typically interested in computing a function based on sensor readings (such as min or average of sensor readings or a more complex function such as finding the most likely location of an object based on sensor readings).
In such networks, an enormous amount of sensor readings is generated but, unfortunately, straightforward approaches for data processing in such large-scale deployments lead to energy-waste and long response-times from sensing to actuation.
This project will rectify that.
Because these networked embedded computers are resource-constrained (typically battery- operated, reduced computing and communication capabilities), energy-efficient operation is important. Additionally, because of the physical interaction, it is often necessary that the delay from sensing until actuation is low and bounded.
In order to achieve these two goals, (i) low delay and (ii) low resource usage, the problem of performing scalable and efficient information processing emerges as being one of the major unsolved problems in large-scale dense cyber-physical systems.
With "efficient information processing" we mean that the desired computation is performed while consuming very little resources (such as energy, communication links, memory and processor). With "scalable", we mean that the consumption of resources increases slowly or not at all as the number of sensor readings to be processed and/or the number of embedded computer nodes increases.
The major novelty of this project is effectively in the co-design of distributed algorithms for sensor data processing and underlying networked distributed computing systems with corresponding resource management schemes such that the utilization of resources is low.
We have recently introduced a family of Medium Access Control (MAC) protocols that are inspired on Dominance (or, Binary-Countdown) protocols. By associating the priorities of messages to physical quantities (such as temperature), several high-performance algorithms for data processing can eventually be devised in which time-complexity is independent of the number of nodes. We denote this simple, but powerful, mechanism as Physical Dynamic Priority Dominance ((PD)^2) protocol. We advocate its use as a key component in sensor applications where it is crucial to compute aggregate quantities with low time complexity, even for very dense systems. (PD)^2 is an example where communication and computation are tightly connected with the physical environment, which is a fundamental feature of CPS.