Sample research projects include resilient and energy-efficient architecture design, design for test and testability, multiprocessor memory systems, internet computing, analysis and optimization of wireless ad hoc networks, green information technologies, adaptive routing for multi-core processors, and reconfigurable computing.
Bioengineering and health-related applications have become an important focus in engineering. Computer Engineering research with a bio- or health-related focus includes research performed by Prof. C.S. Raghavendra on low-power sensors for patient monitoring.
In collaboration with Dr. Ashit Talukder of JPL and USC Health Sciences, Prof. Raghavendra is exploring continuous monitoring of patients with several sensors attached to form a body area network. Measurements are taken from interstitial fluid and with sensors to measure other physiological parameters. These measurements are processed and communicated via customized radio hardware wirelessly to medical personnel. Distributed coordination of these sensors is performed using Markov Decision Process to optimize the energy consumption to meet with life time and quality of health monitoring constraints.
Prof. Murali Annavaram recently started exploring the application of mobile devices in health care. His team received a coveted NIH supplemental grant for two years to build a complete mobile end-to-end system, called KNOWME network, which allows doctors to monitor the behavior of pediatric obesity subjects. They are now in the process of developing software in preparation for the field trials. This work is also funded by Qualcomm and Nokia. Prof. John Granacki, in collaboration with Prof. Ted Berger in Biomedical Engineering, is constructing circuits that model neurons in the hippocampus. The pair's work on "neural chips," which perfectly mimic the behavior of living hippocampus cells, has received significant attention in the press, and has garnered funding from NSF and DARPA to restore lost memory function. An electronic bypass of a damaged hippocampus could restore the ability to create new memories. The project might eventually go further, enhancing normal memory or helping to deduce the particular codes needed for high-¬level cognition.
Prof. Alice C. Parker has received NSF funding to study the use of nanotechnologies like carbon nanotubes in electronic circuits that model neurons in the cortex. Her group is the first group to study nanotechnologies for this purpose, and a portion of an electronic neuron is under construction, using a carbon nanotube in collaboration with Prof. Chongwu Zhou.
The computer engineering group at USC is engaged in various research activities in Green Computing technologies, targeted to using resources in computing systems in an energy efficient and environmentally responsible manner. The common theme of this research is better designs for circuits, system components, system architectures, interconnects and software with the overarching goal of reaching perfect energy/performance proportionality in computing and communication systems. We highlight some of our research projects in the following.
A major focus in computer system architecture is green datacenters. We are developing energy complexity and models for computing and communication within a datacenter, techniques to design, exploit and manage heterogeneous components at all levels of a datacenter (for example, by throttling EPI between sequential and parallel phases of execution), and global/adaptive power and thermal management techniques. We are exploring the disaggregation of datacenters into heterogeneous nodes for better resource sharing and provisioning, and the integration of an entire data center on a chip, using 3-D stacking technology (which exploits emerging technologies such as solid disk and Phase Change Memories). Besides the research on green datacenters, the architecture group is exploring trade-offs between power, performance and reliability at all levels, including memory, caches and cores using new modeling techniques. Parallelism is a major focus. Better, more efficient simulation and modeling techniques are being developed to help design and research future low-power multithreaded systems, including performance, power, energy, and thermal/cooling environment. In one project we demonstrate that complex out-of-order processors are more power/energy efficient than simple cores clocked at higher rate for real-time systems.
Our green network research encompasses the internet, wireless networks, and on-chip interconnection networks. We have recently demonstrated internet router designs with at least an order of magnitude reduction in power consumption over current optimized designs. The “autonomous network research group” explores energy-efficient "green" operation of cellular-based wireless networks. The key idea is to turn off base stations selectively in high density urban environments during off-peak hours while maintaining quality of service for mobile users. A new methodology for designing resource-efficient on-chip networks has been developed. It optimizes cost, performance and power for interconnecting multiple processor cores within a single multi-core chip by exploiting the communication characteristics of applications.
CENG faculty members involved in this effort include Murali Annavaram, Michel Dubois, Rahul Jain, Massoud Pedram, and Viktor Prasanna.
Visit their websites below.
|Murali Annavaram||Super Computing In Pocket (SCIP) Group|
|Peter A. Beerel:||USC Asynchronous CAS/VLSI Group|
|Melvin Breuer:||Test Generation and Validation for Crosstalk in Digital Circuits|
|A Test and Validation System for Crosstalk Induced Errors|
|Kai Hwang:||USC Grid Security Research Group|
||Autonomous Networks Research Group|
|Alice Parker:||BioRC Biomimetic Real-Time Cortex|
|Massoud Pedram:||SPORT: System Power Optimization and Regulation Technology Lab|
|Timothy Pinkston:||SMART: Superior Multiprocessor ARchiTecture Interconnects|
|FPGA/Parallel Computing Group|
|Cloud Computing Group|
|Big Data Analysis Group|
|Smart Grid Group|
|Konstantinos Psounis:||Networked Systems Performance and Design Lab|