Non-linear coding and decoding strategies exploiting spatial correlation in wireless sensor networks

Abstract

The authors consider the acquisition of measurements from a source, representing a physical phenomenon, by\ud means of sensors deployed at different distances, and measuring random variables (r.v.’s) that are correlated with the\ud source output. The acquired values are transmitted over a wireless channel to a sink, where an estimation of the source\ud has to be constructed, according to a given distortion criterion. In the presence of Gaussian random variables (r.v.’s) and\ud a Gaussian vector channel, the authors are seeking optimum real-time joint source-channel encoder–decoder pairs that\ud achieve a distortion sufficiently close to the theoretically optimal one, under a global resource constraint, by activating\ud only a subset of the sensors. The problem is posed in a team decision theoretic framework, and the optimal strategies are\ud approximated by means of neural networks. The analysis investigates the generalisation capabilities of the proposed\ud approach, by showing insights into the structure of the problem. The surprising outcome is that a quasi-static application\ud of the approach reveals to be sufficient to maintain quasi-optimal performance under a dynamic environment (e.g. with\ud respect to nodes’ positions)

Neural Approximations of Analog Joint Source-Channel Coding

“…In order to allow general nonlinear coding-decoding strategies, we consider here the approach of [15], which we briefly summarize. Coding and decoding strategies for the estimation of 4 are of the form:…”

Approximating Optimal Estimation of Time Offset Synchronization With Temperature Variations

“…Remark 3: In [21], [22], and [24] and in this paper, a neural approach is applied to approximate optimal control or state estimation (optimal control in [21] and [24], state estimation in [22] and in this paper). The main difference between those works and the idea presented here relies on the numerical approach used to derive the approximation.…”

“…The main difference between those works and the idea presented here relies on the numerical approach used to derive the approximation. In [21], [22], and [24], the functional cost provides indication in the direction of the optimal solution without the explicit knowledge of it. The minimization process is therefore driven by sampling the cost and its gradient and performing a descent step (also known as stochastic gradient [25]).…”

Linear and non-linear montecarlo approximations of analog joint source-channel coding under generic probability distributions

“…In order to allow non linear codingdecoding strategies, we consider here the approach of [14], which we briefly summarize. Coding and decoding strategies for the estimation of x 4 are of the form:…”

“…The structures of the coders impact on the decoder and viceversa. This is achieved here through their joint numerical optimization; when the regular back-propagation for training neural networks is used, one can note that its initialization forf i (·) depends on the actual values of the derivative of the input ofĝ(·) (e.g., see the appendix of [14]). Apparently, the effect of the non-linearity here (as in all neural functional approximators) stems from the presence of the parameter vectors inside the non-linear activation functions.…”