Equipment Grant Proposal Date:

  March 24, 2003

 

DEPARTMENT:

Department of Electrical Engineering and Computer Sciences

UC Berkeley

 

 

Submitted to Intel Corporation Academic Relations by

 

Professor Shankar S. Sastry

NEC Distinguished Professor of Electrical Engineering and Computer Sciences and Bioengineering
Chairman, Department of EECS 

http://robotics.eecs.berkeley.edu/~sastry/

 

Federal Tax ID No

94-6002123

Email

sastry@eecs.berkeley.edu

Phone

510.642.0253

Fax

510.642.2845

Shipping Address

231 Cory Hall, M/S 1770

University of California

Berkeley , CA    94720-1770

Secondary Contact Name

Ferenc Kovac

Secondary Contact Email

ferenc@eecs.berkeley.edu

Secondary Contact Phone

510.642.6952

 

 

 

 

1.  Overview

 

UCB’s Department of EECS is generally ranked as one of the top EECS departments in the world.  The department has 1,042 majors in the College of Engineering , 244 Computer Science majors in Letters and Science,and 524 graduate students.  In addition, virtually all undergraduates at UC Berkeley pass through one of our classes at some time during their career.

 

 

2.  Need and Opportunity

 

We are in the final stages of planning an EECS Wireless Foundations center, dedicated to providing the theoretical and algorithmic foundations for tomorrow’s wireless systems.  It will be tightly coupled with an instructional component in communication, fundamental RF system design and simulation, and wireless MEMs and devices.  Some of the topics to be covered include propagation and antennas, low power systems, signal processing techniques and implementation, modulation and coding, network architectures, software and protocol performance, sensor and ad-hoc networks, and network security.

We believe that tomorrow's wireless environment will consist of heterogeneous systems operating at many scales.  Many of these systems will be distributed, mobile, and some will incorporate low power, unreliable, and computationally limited components.  The number of different wireless systems competing for the use of scarce spectrum will continue to proliferate with different services having distinct quality of service requirements.

Yesterday's theory and algorithms are not adequate to address this:

Architectures have been primarily driven by a "point-to-point" philosophy -> we need to better understand a network viewpoint wherein nodes can cooperate intelligently taking advantage of the special properties inherent in wireless communication.

Architectures have been primarily centralized -> we need to develop highly distributed architectures and algorithms that are still robust and energy-efficient on a system basis.

Current wireless systems share spectrum through a rigid frequency partition and tight regulatory requirements -> we need more flexible and adaptive spectrum sharing mechanisms that can mostly be self-enforced at the level of the wireless nodes themselves.

Research efforts have been primarily compartmentalized -> we need highly inter-disciplinary research across signal processing, communications, game theory, and networking.

If these are not done soon, we run the risk of having innovation in the future being hobbled by well intentioned but misguided regulatory regimes and legacy systems put in place today.

 

Our research focuses on getting fundamental insights and developing better ways to think about wireless systems.  In the course of doing this, we come up against many previously formulated open problems in network information theory (e.g. non-degraded broadcast channels, relay channels, two-way channels, etc.) that have evaded the community's grasp. We believe that by taking a fresh perspective and formulating the problems in a different way, it is possible to get further understanding. In addition to information theoretic models and bounds, we study protocols, codes, and algorithms.  Some topics we are currently exploring are:

Understanding the 3 R's of spectrum management: spectrum reduce, reuse, and recycling.

Intelligent cooperation between nodes to serve network functions

Understanding the power of feedback and common protocols in enabling spectral efficiency at the network level.

Distributed signal processing for sensor networks.

Exploitation of mobility and ultra wideband to communicate.

Understanding what "quality of service" means for wireless systems.

Fundamentally rethinking the architecture of multimedia-over-wireless systems.

 

3.  Commitment by University

 

We have an excellent professional instructional and research staff that ably administers and maintains our laboratories and have committed infrastructure funds for the center and lab , to be located in Cory Hall.

 

 

4.  Project Requests

 

4.1   Wireless Labs

 

General List of Equipment Needs:

 

42 Dell Precision Workstation 650, Xeon, 3.06GHz, 512K Cache, 19” LCD Monitor (Value:  approx. $160k)

 

                32 Ipaq Pocket PC with Intel X-scale processor (Value: approx. $30k)

 

6 Tablet PCs (Value:  approx. $10k)

 

4 Dell PowerEdge 6600 severs (Terminal Server, Project Space, Video), Xeon 2.0GHz w/2MB Cache, Redundant Power. (Value: approx. $80k).

 

 

Expected users, course(s) covered and numbers of students per academic year:

 

Our students are given a lot of freedom to formulate problems and plans of attack on their own, though the faculty advisors are certainly always available to share ideas and brainstorm with the students. Most of the faculties meet with their students at least once a week, and many students also collaborate with each other on at least some of their research. While doing research is certainly the core part of the student experience here, the following classes will also be served by this grant:

EECS117:  Electromagnetic Fields and Waves

EECS217:  Microwave Circuit Design

EECS221A: Linear Systems

EECS226A: Stochastic Processes, Estimation, and Detection

EECS224: Digital Communications

EECS225x: Signal Processing

EECS229: Information Theory

EECS290-X:  Special Topics Courses on Coding Theory, Wireless, and other advanced topics

 

Location where equipment will be placed:

 

In addition to the Wireless Foundation center on the second floor of Cory Hall, we plan to renovate and combine instrumentation labs with PCs used in design and simulation.  We also plan to build an anechoic chamber for use by the students for antenna experiments.

 

Applications:

 

In addition to RF design and simulation tools such as ADS, we plan to run significantly compute-intensive custom algorithms and simulations.

The Ipaq and Tablet PCs will be a testbed for the development of multimedia applications on the wireless infrastructure landscape.

 

 

New or Replacement?

 

Replacing and augmenting older Pentiums and Sun Workstations

 

Faculty or Lab Managers Responsible:

 

Our core group of faculty:

Venkat Anantharam: Networking & Information theory

Michael Gastpar: Signal processing & information theory

Kannan Ramchandran: Signal processing & communications

Anant Sahai: Information theory & control

David Tse: Information theory & wireless communications

 

Affiliated faculty:

Ali Niknejad

Bob Brodersen

Laurent El Ghaoui

Dave Messerschmitt

Jan Rabaey

Shankar Sastry

J. S. Smith

Ion Stoica

Pravin Varaiya

Jean Walrand

Avideh Zakhor

 

 

4.1A Specific Project:  BASiCS

 

The BASiCS group is involved in a number of research projects:

 

Low Complexity Video Encoding: We are in the process of developing a novel video coder, which directly addresses the problem of uplink video in wireless wide-area networks.  The coding scheme enables low complexity encoding and decoding by pushing most of the computational complexity to the network. It is also more tolerant to channel losses as compared to standard video codecs.  (Details below).

 

Distributed Multimedia Transmission from Multiple Servers: We have developed an algorithm to deliver near-constant quality of video by parallel streaming from multiple servers.

 

Distributed Audio Processing: We address the problem of distributed compression of audio sources for the purposes of acoustic beamforming, speaker location and tracking, blind source separation, 3D sound field capture, etc.

 

Sensor Networks: We are working on algorithms for reducing energy consumption in sensor networks using distributed adaptive signal processing algorithms. The algorithms track and exploit relevant correlation structures of the network.

 

Secure Hashing and Authentication for Image databases:  We are developing scalable algorithms for indexing very large image databases which are robust to malicious attacks on the databases as well as to unauthorized users.

 

 

4.1B Specific Project:  BEAR

 

 The BErkeley AeRobot (BEAR) project is a research effort started in 1996 that encompasses the disciplines of hybrid systems theory, navigation, control, computer vision, networked communication, discrete decision making under uncertainty, coordinated mission planning and and reinforcement learning.  The BEAR vehicle fleet is currently composed of autonomous helicopters (Yamaha and Kyosho radio-control helicopter airframes) and all-terrain vehicles (ActiVision Pioneer ground robots), all equipped with GPS/INS, cameras, on-board real-time controllers and a range of sensors.

 

In the BEAR project, unmanned air vehicle (UAV) navigation sensors are based on an integrated INS/GPS and integrated position and attitude sensors, embedded real-time flight controllers and auxiliary computing systems, wireless communication, and vision capabilities, allowing the UAV to achieve autonomous flight by following dynamically uploaded commands from an external souce.  The flight control system is capable of performing autonomous hover, turns, nose-in maneuvers, forward flight and longitudinal-lateral flight with fixed headings.

 

 

         4.2  Wireless/MEMS/Devices Labs

  

General List of Equipment Needs:

 

16 DellWorkstation 650, Xeon, 2.40GHz, 512K Cache (Value:  approx. $48k).

 

Expected users, course(s) covered and numbers of students per academic year:

 

EE245 (Intro to MEMS Design)

EE143 (Processing and Design of ICs)

EECS299 (Independent Study)

 

Location where equipment will be placed:

 

Cory Hall

 

Applications:

 

MEMS CAD software including SUPREM, SPICE, SAMPLE, SIMPL and MATLAB.  Graduate students, post-docs and visiting industrial research fellows in the device research group for conducting their research, will use the new computers.  These include running various computationally intensive programs, such as MATLAB equation solver programs written for quantum effect simulations and SPICE circuit simulators, device modeling and data processing, graphics manipulation and also comprehensive desktop publishing for conference and journal paper preparation and documentation.  Experimental and simulation data will be stored in these computers.

 

New or Replacement?

 

New and replacement of old Pentium II and Celeron CPUs.

 

Faculty or Lab Mangers Responsible:

 

            Professor Tsu-Jae King

            Engineering Services Group (Ferenc Kovac)

 

 

4.2A Specific Project:  Wireless/MEMS Laboratory

 

EE245 is a new and highly innovative graduate-level interdisciplinary course in Micro Electro Mechanical Systems design and fabrication.  (It is also offered as Mechanical Engineering C218.)  In this course, students will design and fabricate MEMS with applications to wireless sensor networks.

 

The facility needed for this new course will also serve as an instructional laboratory for EE143, in which students now be able to design and fabricate MOS transistors and poly-Si surface microstructures, and execute projects that require extensive CAD software use during the design simulations, and metrics and characterization in the test phase.

 

 

4.2B Specific Project:  Device Research Group

 

The UC-Berkeley semiconductor device research group has a tradition of high impact research.  Examples of contributions to the field include the BSIM MOSFET SPICE model for circuit simulation, and the FinFET transistor structure for ultra-scaled high-performance CMOS technology.  Research at the forefront of device physics and technology requires high-speed computing, for accurate process and device simulation and design optimization.