My previous research mainly focus on Scheduling with partial channel state informaction (CSI), wireless system design with a emphasis the scheduling alogorithms with software defined radio and Interference management in wireless networks. In the following is a summary of the topics I have worked on.
Scheduling with limited CSI feedback
In this research we are considering the scheduling of the downlink of a single base station cellular system, which contains multiple multi-rate channels. It’s well-known that the queue-based Max-Weight scheduling scheme is throughput-optimal, in the sense that it can stabilize all the queues once the arrival rates are in the capacity region of the network. However, the Max-Weight scheduling requires the channel state information of all the users for the scheduling decision, which increase with the number of both channels and users, which is usually fed back via some uplink control channels. And the full feedback would introduce a significant amount of communication overhead on the uplink, which is very impractical. We consider practical scenarios where only a limited fraction of the channel states can be explored. The goals of our research are: i) to characterize the fundamental impact of partial channel state information (CSI) on network throughput and other performance metrics such as delay, and ii) to develop high-performance and low-complexity algorithms that work optimally subject to partial information.
Wireless system design
In the past few years, it has witness the boost of wireless technologies in our everyday life, in which networking plays an increasingly critical role. On one hand, there have been a fruitful of theoretical results on how to increase the network throughput and maintain the Quality of Service; on the other hand, thanks to Moore’s law, new devices and design techniques (e.g., software defined radio and programmable sensor nodes) become available and affordable to support flexible design of various networking schemes. However there’s still a gap between theory and practice due to practical system constraints (e.g., limited buffer and difficult network-wide synchronization) or oversimplified wireless models, which are often ignored in theoretical analysis. In this work, I am interested in leveraging my knowledge with scheduling, using it to continue my work to develop practical protocols for wireless scheduling. On one hand, I want to bring together a rich set of analytical tools drawn from stochastic processes, optimization and control theory and merge them with real-life issues; on the other hand, during the application of the theoretical results, we would like to identify the implementation issues, address their impact on performance (e.g., throughput and delay) and use the theoretical tools for solutions. And by doing this process iteratively, eventually we could get some practical scheduling schemes which fill the gap between theory and practice.
I have modified Hydra code to support OFDMA on USRP, code available upon request.
Also I have co-mentored a senior design project on implementation of an IEEE 802.15.4-based relay network in Fall 2010.
Currently I am working on including backpressure scheduling on sensor nodes (The lack of network-wide synchronization might be an issue in this scenario).
I have modified Hydra code to support OFDMA on USRP, code available upon request.
Also I have co-mentored a senior design project on implementation of an IEEE 802.15.4-based relay network in Fall 2010.
Currently I am working on including backpressure scheduling on sensor nodes (The lack of network-wide synchronization might be an issue in this scenario).
Interference Management in Wireless Networks
Interference has long been considered as the major constraint on the network capacity. In the pioneered work of interference alignment, it shows that it’s possible to overcast all the interference into the same half of signal space and allow each user to use the rest half as interference-free environment, which means the network capacity might scale linearly with the network size. However, the interference is aligned on some assumptions which are difficult to cope with in practical systems (e.g., the availability of global channel state information (CSI), and network-wide synchronization). In recent years, there have been a lot of works on how to reduce the overhead requirement, how to utilize the outdated channel state information and interference alignment for different types of channels. In this research, we are looking for i) practical interference alignment schemes with low communication and/or computation overhead, and ii) we also tries to reveal the interaction of scheduling and underlying interference alignment schemes for the scenario where no CSI or only a small fraction of CSI is available.