EE359 Project Final Report (pdf)

EE359 Project Proposal: A Practical Approach to Bitrate Control in Wireless Mesh Networks using Wireless Network Utility Maximization

Wireless network performance can vary significantly based on selection of parameters like transmission rate and power. Current rate selection schemes like AMRR and SampleRate[1] have shown terrible performance in testbed experiments. It is seen that self-interference causes the rate adaptation schemes to devolve to the lowest rate, thereby reducing performance.

For this project, I will explore estimating the channel from a mixed view of signal-to-noise ratio and packet reception ratio. A high signal-to-noise ratio with a low reception ratio typically means the channel is interference limited, and lowering bitrate may not be favorable for such cases.

As a starting point, I look at theoretical research for ways to maximize performance in mesh networks. Specifically, I look at research in Wireless Network Utility Maximization (WNUM)[2] as a guiding point for implementing a practical rate and power adaptation scheme for wireless mesh networks.

Implementing WNUM for a practical system would require understanding the assumptions inherent in the WNUM model and trying to compensate for those assumptions. Some specific assumptions and their corresponding system implementation workarounds are the following:

• WNUM assumes perfect, immediate feedback for channel state. In existing WiFi systems, the only immediate feedback available to the transmitter is the ACK packet from the receiver. I assume the wireless channel to be symmetric, which means that the SNR at the receiver can be estimated using the SNR of the ACK packet from the receiver. This still makes the channel estimate biased, since the transmitter will only receive ACK packets when its transmissions are successful. That is, transmitter does not get any feedback for failed packets. To compensate for this, I plan to use a packet reception ratio based channel estimate for failed transmissions.

• WNUM ignores interference between multiple flows. This assumption is important for mesh networks, because as said earlier, self and inter-flow interference can lead to very bad channel estimates, thus devolving the network to lowest rate. To circumvent this problem, the implementation will try to filter sudden changes in channel estimates which may be caused due to interference. On the positive side, the average power constraint in WNUM will be beneficial in restricting the inter-flow interference in the network.

I also expect other complications to arise during implementation, that I may not have foreseen yet.

The implementation will be done using the Click modular router[3], which will make the implementation immediately usable for both ns2 simulatons and running on actual hardware. For the purpose of this project, I plan to run most tests in ns2, while also trying to do a proof of concept hardware implementation.

References

1. John Bicket, “Bit-rate Selection in Wireless Networks,” MIT Master’s Thesis, February 2005 (pdf)
   

2. D. ONeill, A.J. Goldsmith, and S.P. Boyd, “Optimizing adaptive modulation in wireless networks via utility maximization,” International Conference on Communications (ICC), 2008, Beijing, PRC. (pdf)
   

3. “Click Modular Router” (website)