Heather (Haitao) Zheng

Associate Professor

Harold Frank Hall, 1121

Dept. Computer Science

University of California

Santa Barbara, California 93106-5110

Phone: +1 805 893-3560

Fax: +1 805 893-8553

imgLeft example

imgLeft example

-->

Computer Science University of california, Santa barbara

Heather Zheng's Research


My current research focuses on designing robust and scalable systems to provide reliable and efficient spectrum usage to wireless networks. It combines multiple disciplines from algorithms, systems, to economics and security.
I am very fortunate to work with a group of talented students at the Link lab.

Current Research Projects


SAFIRE -- Efficient and Reliable Dynamic Spectrum Access
Dynamic spectrum access aims to improve spectrum utilization by networks finding/sharing spectrum dynamically. Yet the reliability of spectrum access is often ignored. Because networks (of various scales) interact and share spectrum dynamically, their spectrum usages flucturate over time. Only if the reliability of dynamic spectrum access is improved, can major wireless networks adopt this new model and provide usable spectrum to innovative wireless technologies and services.
In this project we build, SAFIRE, a robust network architecture for dynamic spectrum access. Our goal is to provide reliable and efficient spectrum usage across wireless networks of various scales, while adapting spectrum usage to traffic demand. SAFIRE achieves reliability by statistically regulating spectrum demand to proactively avoid congestion, and by applying distributed coordination to quickly adapt spectrum usage to network and spectrum dynamics, all while utilizing spectrum efficiently.  More importantly, SAFIRE explores the tradeoff between reliability and spectrum utilization and allows network providers to customize spectrum usage based on their own reliability requirement. Our most recent results include papers at Infocom08, WiCON08 and DySPAN08.
See SAFIRE project website for more details.
The SAFIRE Project is being supported by the National Science Foundation (CNS-0832090) and Samsung.

VERITAS -- Dynamic Spectrum Auctions: Systems and Algorithms
In many cases, spectrum is not free and can be traded between networks/players. To enable dynamic spectrum trading, we propose an "eBay" like open spectrum market that uses on-demand auctions to distribute spectrum matching user demands. We currently focus on two specific problems (1) Clear auctions in real-time; (2) Make auctions economic-robust to market manipulations.  The biggest challenge we face is that the complex interference constraints among bidders make the new auctions fundamentally different from conventional auctions, and many existing economic solutions fail or become highly inefficient.
We have made significant progress alone this line. We have developed fast auction clearing algorithms, and truthful auction designs in both single-sided and double auctions. In single-sided auctions, a seller auctions its spectrum to many smaller players; while in double auctions, multiple sellers and buyers trade spectrum using auctions. Our truthful designs discourage participants/bidders from manipulating their bids because doing so cannot lead to any gain. We also examined the impact of auctions on network performance. Our most recent results can be found from papers at Infocom09, MobiCom08, DySPAN07 and SDR'08. 
See VERITAS project website for more details.
The VERITAS project is being supported by the National Science Foundation (CNS-0915699, CNS-0721961).

LINK Lab's Agile Radio Testbeds
We are developing agile radio testbeds using
(1) 802.11a/b/g cards: We have modified Linux MadWiFi drivers to support TDMA driven MAC protocols with fast channel switch.
(2) USRP GNU Radios: We are implementing our distributed spectrum sharing protocols using GNU Radios.
See the project website for more details.

Proactive Spectrum Access
In the context of opportunistic spectrum access, secondary users (or cognitive radios) must avoid disrupting primary users. Conventional approaches use a reactive "sense and leave" approach, therefore secondary users experience unpredictable interruptions in their spectrum usage. This project aims to reduce such disruptions by proactively predicting future spectrum availability and scheduling spectrum access for long-term stability. In this way, secondary users can fully utilize rich yet dynamical spectrum white-spaces while obtaining usable and predictable spectrum to support network services.
Our main results are a set of history-based prediction algorithms that allow secondary users to choose channels with the longest expected idle periods in near future. This minimizes the amount of channel switches and thus disruptions to ongoing communications.
See the project website for more details.

Past Research projects

Collaborative Decentralized Spectrum Sharing
To utilize spectrum efficiently, wireless networks and their nodes must coexist and share spectrum, forming a large-scale complex system. Spectrum must be distributed dynamically among network nodes to meet their time-varying traffic demands while eliminating interference between nearby peers.  The performance requirements of these dynamic networks means that all management protocols have to be extremely efficient.
We address these challenges by developing a distributed coordination architecture where nodes decide and adapt their spectrum usage by learning environments and coordinating with peers. Upon detecting a sub-optimal spectrum usage, nodes trigger local coordination events and apply collaborative adjustments to approach a globally optimized spectrum allocation. By making mutually consistent decisions that maximize system-wide efficiency, nodes avoid undesired fluctuations and quickly adapt spectrum usage to network dynamics. To build robust systems with different deployment considerations, we proposed two types of coordination formats: (1) an explicit coordination approach where devices negotiate spectrum usage through message exchange, and (2) an implicit coordination approach where devices independently adjust usage following predefined rules without any message exchange.

        - Understanding the Power of Distributed Coordination for Dynamic Spectrum Management, MONET
           October 2008. [ Paper on Explict Coordination ]
        - Distributed Rule-Regulated Spectrum Sharing, JSAC Jan. 2008  [ Paper on Implicit Coordination ]

Spectrum-aware Routing
To achieve reliable multi-hop communications in dynamic spectrum systems, coordination among links alone is insufficient. When portions of the network suffer from critical shortage of spectrum, such as unexpected heavy usages at legacy users or sudden surges of spectrum demand, link coordination and adaptation cannot eliminate disruptions because of the lack of spectrum. My initial study showed that integrating routing and spectrum coordination can double
the throughput compared to the decoupled approach, but suffer from significantly higher complexity. Following this work, we proposed a low-complexity, distributed route selection and channel allocation algorithm to achieve high-throughput multi-hop communications.  By considering link reliability during route selections, nodes can intelligently plan reliable paths away from the problematic areas. In addition, nodes can change route configuration to overcome disruptions from spectrum fluctuations and maintain reliability. 
In parallel, I also researched on low-complexity routing solutions for QoS support. We proposed QUORUM (QUality Of service RoUting in wireless Mesh networks), a routing protocol optimized for WMNs that addresses QoS in mesh networks. QUORUM integrates a novel end-to-end packet delay estimation mechanism with stability-aware routing policies, allowing it to more accurately follow QoS requirements while minimizing misbehavior of selfish nodes.

        - High Throughput Spectrum-aware Routing for Cognitive Radio Based Ad-hoc Networks, CROWNCOM 2008.
           [ Paper on Spectrum-aware routing]
        - QUORUM - QUality Of service RoUting in wireless Mesh networks, ACM MONET, December 2007. [PDF]

Integrating Spectrum Allocation with Collaborative Relay
Coordinated spectrum sharing can be combined with other collaboration techniques. We start from collaborative relays that use node relaying to achieve spatial diversity without requiring physical antenna arrays at end-devices.  While many
studies have demonstrated its effectiveness in an isolated single source-destination system, applying cooperative relays to large-scale wireless networks remains challenging. In TMC08, we show that a large wireless system with cooperative relays can be penalized be the elevated level of interference it produces, and the design of relays must be integrated with spectrum allocation. By exploiting the inter-dependency between spectrum allocation and cooperative relay, we model the penalty by an increase in spectrum resource usage and translate it into a penalty in the single-link throughput. This throughput penalty serves as a reference for designing and evaluating collaborative relays in small, isolated environments.

         - Understanding the Impact of Interference on Collaborative Relays, IEEE Transactions on Mobile Computing,
           Vol.  7, No. 6, pp 724-736, June 2008. [TMC08]

Research Projects Before UCSB


At Wireless and Networking Group, Microsoft Research Asia (04/04-07/05), I initiated the following projects on open spectrum systems:

At Wireless Research Laboratory, Bell-Labs, Lucent Technolgies (08/99-03/04), my research was on radio resource allocation for broadband wireless networks:

My PhD thesis research at Univ. of Maryland College Park was on multimedia communications, a cross layer design framework to provide resource efficient multimedia delivery over noisy medium: