A learning-based algorithm for optimal mac parameters setting in IEEE 802.15.4 wireless sensor networks

Abstract

Recent studies have shown that the IEEE 802.15.4 MAC protocol may suffer from severe limitations, in terms of reliability and energy efficiency, if CSMA/CA parameter settings are not appropriate. On the other hand, choosing the optimal setting that guarantees the application requirements with minimum energy consumption, may not be a trivial task in a real environment, where the operating conditions change over time. In this paper we propose a LEarning-based Adaptive Parameter tuning (LEAP) algorithm that, in addition to adapting the CSMA/CA parameter settings to the time-variant operating conditions, also exploits the past history to figure out the most appropriate settings for the current conditions. Simulation results show that, in stationary conditions, the performance of the proposed algorithm is very close to an ideal (but unfeasible) algorithm. It is shown that LEAP is able to select in dynamic scenarios the optimal settings faster than related algorithms.

Intelligence for Embedded Systems

“…Here, adaptation is needed at the communication level, e.g., to adapt online the parameters of the protocol or the physical communication [234,235] or grant clock synchronization among units. Clearly, the possibility to exchange information among embedded units introduce another level of potential processing which can be either distributed or hierarchical (controlled process, cognitive mechanisms).…”

A reconfigurable and element-wise ICI-based change-detection test for streaming data

“…On the other hand, the window-wise processing induces a structural detection delay, since changes can be detected with window coarseness rather than at each single sample. Such a delay might not be acceptable when the change magnitude is large, and might represent a problem in those scenarios where the detection delay is a critical figure of merit such as the detection of a contaminant in an intelligent building [14] or the detection of variations in the topology or in the working conditions of a sensor network [15].…”

Distributed traffic estimation and throughput optimization in random access networks

“…Typically, performance is optimized by choosing the CSMA parameters such as binary exponents, backoff limit and retry limit. A representative optimization problem can be formulated as follows [65,72]:…”

“…For both the accurate and simplified models, the performance heavily depends on the accuracy of the underlying models [72]. Moreover, as discussed in Section 2.3, most models require certain parameters' values such as the number of nodes in the network, which is normally unavailable in a distributed network.…”

“…Specifically, ADAPT algorithm increases or decreases the CSMA/CA parameters based on local measurements of packet success rate. The ADAPT does not require prior information on network and can be practically used for any network scenarios [58,72]. However, the ADAPT has the issue of oscillating between adjacent parameters and never stabilizing the system under a given constraint.…”