Date of Award


Document Type

Campus Access Dissertation


Computer Science and Engineering

First Advisor

Srihari Nelakuditi


Information redundancy exists in every aspect of our lives. In some cases, it is created in purpose to resist the possible failure under unreliable environments. For example, backing up important data, or the checksum of network packets; But in many other cases, information redundancy is created unconsciously. Same news could be heard from different friends, just like a same data packet could be received from different neighbors in a wireless network, or a same event could be sensed by different sensors in a mobile sensing application. As the opposite side of efficiency, unconscious redundancy has normally been considered as useless, or even harmful. In this work, we try to disprove the traditional argument in wireless environments, by showing that different wireless protocols/applications can be built on top of redundancy to improve network throughputs/enable interesting applications. The examples lie from PHY layers to application layer, and cross domains including both wireless networking and mobile computing.

At PHY/MAC layer, we observe that many collisions are caused by redundant transmissions. We argue that when one of the colliding packets is previously overheard, its interference can be canceled to decode the other packet. In other words, when a receiver overhears a packet, it becomes effectively immune to the interference caused by the packet's subsequent transmission. We refer to this as known interference cancellation (KIC). We identify the scenarios in which KIC is applicable, then implement KIC on USRP/GnuRadio testbed to demonstrate its feasibility and conduct QualNet simulations to illustrate its potential performance gain.

At MAC/network layer, wireless network coding has been shown to reduce the number of transmissions needed by exploiting the redundancy caused by the broadcast nature of the wireless medium. In our work, we first present IMIX, an intra-flow wireless network coding scheme which has the potential to save transmissions and therefore improve network throughput. We show that the coding gains with IMIX can be further improved by selecting routes with the awareness of overhearing opportunities. We then propose I2MIX that integrates inter- and intra-flow wireless network coding to extracts benefits from both. We validate I2MIX through Roofnet trace based evaluations and show that it can reduce the total number of transmissions.

At application layer, we envision TagSense, a mobile phone based auto-tagging system that can automatically sense and tag the people/activity/context in a picture, The main challenge pertains to discriminating phone users that are in the picture, from those that are not. We proposed method by exploiting the redundancy from different sensing domains. We prototype TagSense on 8 Android phones, and demonstrate its effectiveness through 200 pictures, taken in various social settings. While research in face recognition continues to improve image tagging capabilities, TagSense is an attempt to embrace additional dimensions of sensing towards this end goal.