iMASKO: A Genetic Algorithm Based Optimization Framework for Wireless Sensor Networks

Nanhao Zhu, Ian O'Connor
2013 Journal of Sensor and Actuator Networks  
In this paper we present the design and implementation of a generic GA-based optimization framework iMASKO (iNL@MATLAB Genetic Algorithm-based Sensor NetworK Optimizer) to optimize the performance metrics of wireless sensor networks. Due to the global search property of genetic algorithms, the framework is able to automatically and quickly fine tune hundreds of possible solutions for the given task to find the best suitable tradeoff. We test and evaluate the framework by using it to explore a
more » ... stemC-based simulation process to tune the configuration of the unslotted CSMA/CA algorithm of IEEE 802.15.4, aiming to discover the most available tradeoff solutions for the required performance metrics. In particular, in the test cases different sensor node platforms are under investigation. A weighted sum based cost function is used to measure the optimization effectiveness and capability of the framework. In the meantime, another experiment is performed to test the framework's optimization characteristic in multi-scenario and multi-objectives conditions. Introduction With the widespread development of embedded systems and various wireless communication technologies, wireless sensor networks (WSNs) have gained the attention of industrial and research OPEN ACCESS J. Sens. Actuator Netw. 2013, 2 676 groups all over the world in recent years. The integration of sensing, data processing, and over-the-air transmission into a single miniaturized device enables the deployment of wireless sensor networks in many fields of applications. However, the weak computation ability, limited storage, short communication range, and severe energy constraints, to some extent, limit their use. Therefore, carrying out optimizations on these elements is very useful for the improvement of network lifetime, the reduction of packet loss, end-to-end delay, and other related metrics, all of which are essential to guarantee adequate performance for specific applications. Typically, two classes of objects are optimized in the design of wireless sensor networks: hardware and software. From the hardware perspective, better energy efficiency can be achieved by optimizing the power consumption of related hardware components. By employing an ultra-low power based microcontroller (e.g., MSP430) and configuring it to five different low power modes, a significant amount of energy can be saved. Besides, the reduction of sensing tasks and simplification of data processing algorithms can also be helpful. For the transceiver, some factors, such as the modulation scheme, transceiver packet frame, and duty cycle, can affect power consumption. The use of a high data rate as described in [1] , has proven to be, not only a very energy-effective choice, but also a strategy that greatly improves network reliability. In recent years, some emerging radio-based technologies such as Bluetooth low energy technology [2], Ultra-wideband (UWB) [3], and ANT [4] have also provided excellent choices in lifetime improvement, reliability enhancement, as well as short network latency. From the viewpoint of energy supply, as most sensor motes are battery-driven, the battery type, capacity, and size play an important role in the node lifetime, cost, weight, and deployment ability. Emerging energy harvesting technologies [5] are also highly useful methods for overall energy optimization and power management. In addition, some hardware based algorithms have been proposed for optimizing energy consumption. An adaptive power control algorithm is presented and implemented in [6], by automatically configuring the programmable output power on a transceiver chip according to the distance information between nodes. This algorithm is tested via experiment for the validation of its energy-efficiency in increasing node lifetime. A Dynamic Voltage Scaling (DVS) algorithm [7] applied to microprocessors can also minimize the power consumption by dynamically scaling the supply voltage to match the required performance level. Finally, in the Dynamic Power Management (DPM) algorithm [8], a more traditional method is used by selectively turning off idle state components to save energy. From the software perspective, the optimization method can be grouped into three categories: the development of new communication protocols (MAC and routing) for optimization, the adoption of energy-aware strategies for optimization, and the configuration/exploration of the optimal set of existing protocols for optimization. Firstly, building on the contention-based scheme for collision avoidance and reliable transmission, both S-MAC [9] and T-MAC [10] are proposed to synchronize communication schedules and listening periods to minimize latency, while reducing energy consumption by turning off the radios during sleep periods. Through the use of low-power listening approaches, WiseMAC [11] and B-MAC [12] can save more energy from idle listening. For most of the new proposed routing protocols, optimizations focus on how to select the shortest path for energy saving, while, at the same time, guaranteeing the reliability of the network by reducing the number of communication hops. Secondly, the uses of strategies for optimization include in-network processing, data aggregation, and cross-layer related optimization methods. In-network processing in wireless
doi:10.3390/jsan2040675 fatcat:jf4mf3dfhrgmtpfubut5wy7z5q