Wireless system performance is known to be highly dependent upon the characteristics of the environment. Despite the increasing ability of wireless devices to sense their surroundings, wireless systems have yet to fully leverage contextual data to improve performance. To this end, the Dallas-ARea Testbed for Context-Aware, Cognitive Research (DART-CARs) allows the study of wireless performance in a broad class of mobile and static environments from indoor labs to outdoor high-way speeds within a single testbed. Our hardware platform functions across many different wireless bands to enable real-time, multi-band operation. Such a first-of-its-kind infrastructure is critical for designing context-aware and cognitive algorithms that utilize multiple frequency bands to adapt to dynamic environmental settings. Our project has three major initiatives:
-Deploy a wireless research testbed to enable study across a vast array of urban environments -Design multi-band wireless hardware and reconfigurable embedded software tools that enable context-driven, cognitive switching across frequency bands to improve performance in the aforementioned environments -Develop an open-access data repository of wireless performance in multiple urban scenarios and across frequency bands
The project incorporates topics across disciplines such as wireless theory, embedded programming, hardware design, and ubiquitous computing. Student involvement in courses and research will allow cross-disciplinary intellectual growth. Further, field measurements will be made available to the research community. Finally, handheld mobile devices are increasingly being used as the primary method for under-represented groups to access the Internet. This infrastructure and resulting research will improve the performance these devices.
Wireless system performance is known to be highly dependent upon the characteristics of the environment. Despite the increasing ability of wireless devices to sense their surroundings, wireless systems have yet to fully leverage contextual data to improve performance. To this end, DART-CARs is an NSF-funded infrastructure (NSF CISE CRI) that allows the study of wireless performance in a broad class of mobile and static environments in and around the Dallas area from indoor labs to outdoor high-way speeds within a single testbed. Our hardware platform functions across many different wireless bands to enable real-time, multi-band operation. Such a first-of-its-kind infrastructure is critical for designing context-aware and cognitive algorithms that utilize multiple frequency bands to adapt to dynamic environmental settings. Hence, the major goal of the project is to establish the infrastructure to leverage an awareness of a wireless transmitters contexual settings and improve wireless performance. To define a "context," a channel type is defined as a wireless channel that shows unique link level performance from other types of environmental settings. The channel type, velocity, and signal quality is then directly used to adapt modulation, coding, and frequency band used in order to optimize the througput of a given sender to receiver pair. The infrastructure consists of a channel emulator for repeatable, controllable wireless channels, antennas deployed on campus buses and buildings for continual drive testing, and multiband spectrum analyzers, signal generators, and antennas so that students at all levels can extensively evaluate context-aware protocols on programmable platforms. Intellectual Merit: There are two key thrusts for outcomes on this project. First, an infrastructure was built for the study of context aware wireless protocols which adapt links across transmission modes and carrier frequencies with the following components: (i) a channel emulator for repeatable, controllable channels, (ii) peripheral equipment for examining and generating spectral activity both in the lab and in the field, (iii) antennas which span 450 MHz to 6 GHz deployed on buses and buildings on campus for measurements across multiple environmental contexts, (iv) programmable custom and off-the-shelf hardware for in-lab and in-field measurement and analysis, and (v) Android-based measurement infrastructure which has acquired approximately 200 million measurements for the purpose of examining user mobility patterns and wireless performance across cellular and WiFi networks in diverse contexts across around the world. Second, a key thrust for outcomes in this project was context aware wireless research which included: (i) wireless channel classification and inference, (ii) leveraging context information to build a decision structure for rate adaptation, (iii) training and updating the decision structure on the fly in the field, (iv) multiband link adaptation, and (v) exploiting sequences or expunging anomalies within the data to improve the rate-adaptation decision structure. To disseminate these ideas, we have published 11 conference papers and 2 journal papers. Furthermore, we have filed for a provisional patent based on context-aware adaptation and geometry-based channel inference as well as two invention disclosures. Finally, we have released a video of our infrastructure and a data set for researchers to use. Broader Impacts: The grand vision for the project is to enable devices to train in an on-line manner for context-aware wireless performance improvements (particularly related to link and/or rate adaptation). We have laid the foundation that shows that off-line and on- line methods based on decision trees coupled with algorithms that infer channel types based on some sample measurements in the environment, leads to large gains. Over the course of the project, five undergraduate students (one being Hispanic) and six graduate studnets (two being female) were involved with research. Professor Camp has designed a new course at SMU called Mobile Phone Embedded Design that focuses on designing applications for smartphones that will allow wireless and sensing tasks while preserving the energy of the device. Over the course of the project, the course has been taught three times and there have been multiple group projects from the course that focus on developing the outdoor measurement system. Finally, the mobile phone application that we designed for collecting wireless measurements, has been freely offered on the Android Market since April of 2011. This has allowed users (even if they do not opt into the research) an application that allows them to see the activity level of the wireless channels that they are using and alter their channel accordingly.
