Cooperation in the Low Power Regime: Capacity and Coding
April 11, 2006 11:00 AM to 11:30 AM
Speakers:
Host-Madsen, Anders
|
 |
Abstract: |
In this presentation we consider cooperative communications in the low power regime. The low power regime can be defined as the limit where the power spectral density of the transmitted signal goes to zero, which can be achieved by either letting the bandwidth converge towards infinity or letting the rate converge towards zero. There are a number of advantages to communications in the low power regime: in a non-cooperative channel the energy consumption (energy per bit) is minimized in the low power regime, the communication is interference-free, the modulation is simple, and the communication is covert. In the cooperative channel, energy per bit is not necessarily minimized in the low power regime, but it is close to the minimum, as will be shown, and the other advantages remain. This makes low-power communications ideal for applications like sensor networks. In the presentation, we develop capacity for low power communications for a number of channels. We also consider coding methods that are close to capacity. In particular, we consider two block Markov coding schemes, namely, multiplexed coding and superposition coding schemes. The multiplexed coding using fully multiplexed (FMP) codes, which outperforms the superposition coding theoretically, however, is difficult to implement with practical error-correction codes. We then introduce a partially multiplexed (PMP) coding scheme for the coderate R < 1/2, and propose a PMP code design method using irregular repeat accumulate (IRA) codes. The design method can be extended to build the low-rate PMP codes using irregular repeat zigzag Hadamard (IRZH) codes. The outage analysis shows that the cooperation schemes provide significant gains over the non-cooperative MAC in low-powerregime. Also, the PMP-IRZH coding performs only 0.5 dB away from low-power outage capacity for multiplexed coding.
Joint work with Zigui Yang (Hawaii), Guosen Yue (NEC Labs), and Xiaodong Wang (Columbia). |
|
|
