SOFTWARE: ASP.NET | VB.NET | C#.NET | RAZOR MVC 4 ASP.NET | RESTful Web services
The inherent limitation in energy resources and computational power for sensor nodes in a Wireless Sensor Network, poses the challenge of extending the lifetime of these networks. Since radio communication is the dominant energy consuming activity, most presented approaches focused on reduc-ing the number of data transmitted to the central workstation. This can be achieved by deploying both on the workstation and the sensor node a synchronized prediction model capable of forecasting future values. Thus, enabling the sensor node to transmit only the values that surpasses a predefined error threshold. This mechanism offers a decrease in the cost of transmission energy for a price of an increase in the cost of computational energy. Therefore, finding the right trade-off between complexity and efficiency is very important to achieve optimal results. In this paper, we present a novel data reduction method that outperforms other state of the art data reduction approaches. We demonstrated the efficiency of our algorithm using simulation on real-world data sets collected at our laboratory. The obtained results show that our method was able to achieve a data suppression ratio ranging between 93:2% and 99:8%
we present our new ADRM method, which aims to minimize the amount of data transmitted by each sensor to the Sink
This method is also very robust and is significantly more efficient in term of prediction accuracy and data reduction compared with other state of the art approaches.
This method is efficient since time series data like the ones found in WSN change smoothly over time. And the values of the measurements at neighboring time ticks only change slightly. Moreover, such data tend to follow a specific trend in a cyclic manner.
For instance, when measuring temperature outdoor, the latter increases rapidly in the morning when the sun rises, remains stable at noon, then it decreases gradually when the sun begins to set.
a rectification value is used to adapt the change rate CR in order to extend the model’s prediction horizon. In this section, we will explain how B is calculated automatically rather than being fixed. The shorter the prediction horizon (the less is the number of accurate predictions outputted by the prediction model between two consecutive adjustments) the faster is the estimated CR deviating from the real change rate of collected measurements.
When a readjustment is needed, both the Sink and the sensor calculates the error between the real transmitted measurement and the last prediction that exceeded emax and triggered the transmission.
A Wireless Sensor Network (WSN) consists of small and simple computing devices that are usually in the form of a micro-controller, and have limited computational and energy resources. These devices are powered by irreplaceable batteries since they are typically deployed in harsh or hostile environments. Thus, developing a mechanism that allows such networks to work for a prolonged period of time while oper-ating solely on limited amount of energy is a crucial need for sensor network applications to succeed.
Several approaches aiming to adapt with the severe power constraints have been proposed in the literature. It includes energy-aware protocols and algorithms that have been imple-mented on the physical layer, up to the application layer of the sensor network protocol stack. Since radio communica-tion (transmitting and receiving) is generally the main factor responsible for draining the energy reserves of the sensor node, one should try to minimize the amount of information transmitted by the sensor. Consequently, reducing the number of packets received by intermediate nodes participating in the forwarding of these packets to the workstation (Sink).
we present a new data transmission reduction method, that has a constant complexity and requires a very small memory footprint. This method is also very robust and is significantly more efficient in term of prediction accuracy and data reduction compared with other state of the art approaches. Thus, our approach is challenging the notion of the more complex, the higher the data SR of the prediction model is.
Furthermore, the Sink is running multiple instances of prediction models deployed on a large number of sensor node each. Thus, for a Sink to manage all sensor nodes at the same time, predictions should be produced in the shortest delay possible and should require a very small number of operations. Otherwise, the computational and memory resources of the Sink would be drained rapidly,rendering the latter incapable of managing a large number of models simultaneously.
In this paper, we have proposed a new data reduction method that is fully autonomous and requires no calibration or intervention from the user, only the tolerated error threshold is specified and the rest is automatically adapted according to the collected data. Moreover, despite being very light in term of complexity and memory space, it is proven to be robust and extremely efficient in term of transmission reduction.
Further-more, it has been demonstrated that our method outperforms other state of the art data reduction approaches. This has been proven through an appropriate simulation on real collected measurements of different environmental features. For future work we aim to implement ADRM on real sensor nodes in order to verify the efficiency of our proposal in a real world deployment.