Time-sensitive Information Communication, Sensing, and Computing in Cyber-Physical Systems

Cyber-Physical Systems (CPSs) are increasingly important and pervasive with the convergence of the cyber and physical worlds via digital information. In CPSs, communication, sensing, and computing are three key operations on digital information. These operations are usually time-sensitive and have various temporal requirements. These requirements could be accomplishing a job after sufficient time passes, speeding up a job with "best effort," or finishing a job within a specific time bound. In this dissertation, we study the problems of time sensitive communication, sensing, and computing in CPSs.;First, we study temporally controllable wireless communication. We propose TurfCast, a novel information dissemination service that selectively broadcasts messages based on receivers' lingering time. Only those who have stayed long enough can receive the messages. To implement TurfCast, we propose TurfCode, a nested 0-1 fountain code that enables the broadcaster to transmit either all information or none at all. We extend TurfCast to support spatial "turfs"; and propose TurfBurst, which exploits the Shannon bound to differentiate among receivers based on spatial areas.;Second, we study fast electronic and visual sensing for human localization. It is observed that the presence of human bodies introduces heavy interference to wireless signals. This has been a major cause of inaccurate wireless localization of humans. Wireless signals gathered over a longer period are aggregated to mitigate such interference. We propose EV-Human for fast and accurate human localization. In EV-Human, we propose using video cameras to help estimate human body interference on mobile devices' signals. We combine human orientation detection and human/phone/AP relative position inference to better estimate how a human affects wireless signals. We have also developed a signal distortion compensation model.;Third, we study time-bounded distributed computing on large visual data. Visual data such as images and videos are increasing rapidly in volume as public video camera deployments grow. However, current distributed computing over large datasets with tools like MapReduce favors textual data and cannot cope well with visual data. We design MaReV, an augmented MapReduce framework for big visual data processing with time constraints. In MaReV, we propose an iterative, adaptive, and stratified sampling algorithm for aggregation queries. Our algorithm considers data biases on both values and time costs. In addition, we propose a computation time based load balancing mechanism to accelerate job execution. We also design a profiling module to collect fine-grained runtime statistics and a caching module to facilitate subsequent jobs.;All three of these problems originate from real-world CPS applications and have important practical significance. Simultaneously interacting with the physical world and meeting temporal requirements is non-trivial. Our proposed solutions are novel and span the disciplines of wireless communication, signal processing, computer vision, and distributed systems. We have implemented the above three systems and extensively tested them in real-world experiments. Evaluation results show the effectiveness and time-efficiency of our systems and demonstrate the potential of our solutions for more general time-sensitive CPS application scenarios.