Analytic models and distributed robotics applications for mobile ad hoc networks.

摘要

In this thesis, we introduce new analytic models to study node link stability for realistic wireless mobile network applications by calculating the exact link failure and creation probabilities. These models divide a geographic area into logical cells, where each node roams into one of its neighboring cells by following the discrete-time random walk mobility model. We calculate the probability distribution for a wireless link to be available between two mobile nodes. We derive a new two-dimensional Markov chain whose states represent a node's degree and its number of link failures. We can thus compute two important metrics characterizing the dynamics of a node's random movement: the expected times for the number of link changes to drop below a failure threshold, and for a node's degree to exceed a degree threshold. Because the model is capable of computing the dynamics and the expected value of the number of a node's neighbors, it can be used for modeling various realistic applications including virtual backbone and clustering stability, and estimating interference levels in mobile ad hoc networks (MANETs). Our modeling framework can be further extended to derive a number of additional important metrics to characterize network connectivity, capacity, and survivability.; We also developed two real-life applications of our analytic models: Controlled Dissemination Filter (CDF) for MANETs and a testbed implementation for the Dynamic Survivable Resource Pooling (DSRP) concept in distributed robotics systems. We implemented the CDF framework, together with its architecture and protocols, and showed significant performance improvements related to the CDF without jeopardizing the network and application performance for different dissemination scenarios. We also implemented the DSRP mechanism in our FPGA-based distributed robotics testbed and demonstrated its effectiveness in a search-and-rescue scenario, where robots cooperate to provide the rescuer with a set of pictures to build a panoramic view of the disaster site. The measurements collected from the testbed and the numerical results obtained from the analytic models confirm that the DSRP framework significantly improve the reliability of distributed robotics systems.