This paper presents a joint source and networkcoding scheme, called compressive network coding (CNC), for approximate data gathering in wireless sensor networks. Injecting the concept of compressive sensing into netw...
详细信息
ISBN:
(纸本)9781424492688
This paper presents a joint source and networkcoding scheme, called compressive network coding (CNC), for approximate data gathering in wireless sensor networks. Injecting the concept of compressive sensing into networkcoding avoids the all-or-nothing effect of network decoding, allowing CNC to achieve graceful degradation in data precisions when the energy budget is not sufficient. Based on the observation that sensor links are highly unreliable, we formulate the energy constrained multi-source, multi-hop, and multi-path transmission as a network utility maximization (NUM) problem. A practical distributed algorithm is developed to achieve the optimal utility. We carry out simulations over real sensor data. Results show that CNC consistently outperforms conventional networkcoding, achieving an average gain over 3.7 dB in data reconstruction PSNR. Furthermore, CNC achieves over 10 dB gain for five sensors (out of 53) whose readings contain abrupt changes.
暂无评论