Error-Resilient and Complexity-Constrained Distributed Coding for Large Scale Sensor Networks Kumar Viswanatha, Sharadh Ramaswamy ∗ , Ankur Saxena † and Kenneth Rose ‡ University of California - Santa Barbara Santa Barbara, CA - 93106-9560, USA {kumar,rsharadh,ankur,rose}@ece.ucsb.edu ABSTRACT There has been considerable interest in distributed source coding within the compression and sensor network research communities in recent years, primarily due to its potential contributions to low-power sensor networks. However, two major obstacles pose an existential threat on practical de- ployment of such techniques in real world sensor networks, namely, the exponential growth of decoding complexity with network size and coding rates, and the critical requirement for error-resilience given the severe channel conditions in many wireless sensor networks. Motivated by these chal- lenges, this paper proposes a novel, unified approach for large scale, error-resilient distributed source coding, based on an optimally designed classifier-based decoding frame- work, where the design explicitly controls the decoding com- plexity. We also present a deterministic annealing (DA) based global optimization algorithm for the design due to the highly non-convex nature of the cost function, which further enhances the performance over basic greedy itera- tive descent technique. Simulation results on data, both synthetic and from real sensor networks, provide strong ev- idence that the approach opens the door to practical de- ployment of distributed coding in large sensor networks. It not only yields substantial gains in terms of overall distor- tion, compared to other state-of-the-art techniques, but also demonstrates how its decoder naturally scales to large net- works while constraining the complexity, thereby enabling performance gains that increase with network size. Categories and Subject Descriptors E.4 [Coding and information theory]: Data compaction and compression, Error control codes; G.3 [Probability ‡ This work was supported by the NSF under grants CCF- 0728986, CCF-1016861 and CCF-1118075. ∗ S. Ramasway was with Mayachitra, Inc., USA, at the time of this work. † A. Saxena is now with Samsung Telecommunications Amer- ica, 1301 E. Lookout Dr., Richardson, TX, USA - 75082. and statistics]: Probabilistic algorithms; I.4.2 [Compression]: Approximate methods General Terms Algorithms, Theory, Experimentation Keywords Distributed source-channel coding, Large scale sensor net- works, Error resilient coding 1. INTRODUCTION AND MOTIVATION Sensor networks have gained immense importance in re- cent years, both in the research community as well as in the industry, mainly due to their practicability in numer- ous applications. Sensors are typically low power devices and minimizing the number of transmissions is one of the primary objectives for a system designer. It is widely ac- cepted that exploiting inter-sensor correlations to compress information is an important paradigm for such energy effi- cient sensor networks. The problem of encoding correlated sources in a network has conventionally been tackled in the literature from two different directions. The first approach is based on ‘in-network compression’ wherein the compres- sion is performed at intermediate nodes along the route to the sink [8]. Such techniques tend to be typically wasteful in resources at all-but the last hop of the sensor network. The second approach involves ‘distributed source coding’ (DSC) wherein the correlations are exploited before transmission at each sensor [3]. The basic DSC setting involves multiple correlated sources (e.g., data collected by a number of spatially distributed sensors) which need to be transmitted from different lo- cations to a central data collection unit/sink. The main objective of DSC is to exploit inter-source correlations de- spite the fact that each sensor source is encoded without access to other sources (see Fig. 1). The only informa- tion available before designing DSC is their joint statistics (e.g., a training dataset). Today the research in DSC can be categorized into two broad camps. First approach derives its principles from channel coding, wherein block encoding techniques are used to exploit correlation [1, 9, 18]. While these techniques are efficient in achieving good compression and error-resilience (using efficient forward error correcting codes), they suffer from significant delays and high encoding complexities, which make them unsuitable for several sen- sor network applications. The second approach is based on Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. IPSN’12, April 16–20, 2012, Beijing, China. Copyright 2012 ACM 978-1-4503-1227-1/12/04...$10.00.