^{*}

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (

When a mobile wireless sensor is moving along heterogeneous wireless sensor networks, it can be under the coverage of more than one network many times. In these situations, the Vertical Handoff process can happen, where the mobile sensor decides to change its connection from a network to the best network among the available ones according to their quality of service characteristics. A fitness function is used for the handoff decision, being desirable to minimize it. This is an optimization problem which consists of the adjustment of a set of weights for the quality of service. Solving this problem efficiently is relevant to heterogeneous wireless sensor networks in many advanced applications. Numerous works can be found in the literature dealing with the vertical handoff decision, although they all suffer from the same shortfall: a non-comparable efficiency. Therefore, the aim of this work is twofold: first, to develop a fast decision algorithm that explores the entire space of possible combinations of weights, searching that one that minimizes the fitness function; and second, to design and implement a system on chip architecture based on reconfigurable hardware and embedded processors to achieve several goals necessary for competitive mobile terminals: good performance, low power consumption, low economic cost, and small area integration.

The use of Wireless Sensor Networks (WSN) has increased substantially in the last years [

There are many cases where mobility is an important feature to be taken into account in WSNs. For example, the position of mobile sensors in a WSN must be determined because the performance of event detection and tracking highly depends on the exact location information of the events that must be reported along with the event features [

When a mobile wireless sensor (MS) can link to more than one router of the same network, we can establish a similarity to the well-known procedure of Horizontal Handoff or horizontal handover (also known as intrasystem handoff or handover) in general wireless networks. Horizontal handoff (HH) involves the change of the link of a mobile node to other wireless access points using the same technology. Nevertheless, the increasing complexity of the WSNs and their associated technologies forces us to consider heterogeneous scenarios where more than one network are present and where the nodes can be of different natures, technologies and behaviors. In these cases, and following with the previous similarity, we can follow a scheme of Vertical Handoff (VH) where the mobile node can link to other wireless access points of the same or different sensor networks and technologies.

In this work we study the more general and complex case: the VH process of a mobile wireless sensor moving along heterogeneous wireless sensor networks. In order to explain better the framework of our study, let us consider an example next.

Let us suppose a mobile wireless sensor moving through this scenario (

Let us explain the VH process more in detail. When the MS changes its connection from a router (access point to the network) to another one belonging to a different network, a VH can happen; if both routers are of the same network, it is a HH [

In traditional handoffs only signal strength and channel availability are considered. In the new generation networks new metrics have been proposed [

Some references use the term “cost function” [

We use the fitness function described in

If to larger

The decision phase is the frame for the formulation of the optimization problem. The choice of a network to perform VH is a very important issue when considering QoS parameters. In this choice the key is the weight tuning, because the fitness function is very sensitive to the values of the weights. In this sense, our objective is to find an optimal solution, where each solution is a sequence of weights determining the QoS. Some techniques have been developed to adjust the weights in order to find the minimum fitness,

We propose an embedded architecture that answers to the current trends in mobile computing: low cost, low power consumption and good performance. The HiPEAC Network of Excellence on High Performance and Embedded Architecture and Compilation specifies this trend: “

The low cost, low power consumption and good performance (a fast obtaining of high-precision solutions) are requirements in a small electronic device (for mobile sensor purposes) which moved us to design a custom embedded microprocessor able in reconfigurable hardware to suit the algorithm.

The MS must dynamically choose the best access network for each application flow, so the vertical handoff middleware chooses the access network according to the applications requirements, the user’s preferences and the QoS parameters of the networks. Dynamic scenarios imply rapid processing, where the configurable embedded processors have demonstrated to be a good solution [

In this section we expose the main developments done and results obtained in two fronts: software (a QoS-based decision algorithm for VH) and hardware (the implementation of the algorithm in an embedded microprocessor based on reconfigurable devices).

We name the algorithm we have developed for the VH decision phase taking into account the QoS characteristics of the networks SEFI (from “Weights Combinations Fast

The generation of these combinations is done from a given precision value (h) and two limits determined by the user, WMIN and WMAX, where WMIN < h < WMAX, WMIN > 0 and WMAX < 1. These limits depend of the user’s profile, in other works, the purpose of the application of the mobile device in WSN. The algorithm has been programmed to perform a fast search thanks to some recursive functions. In the computation of the exhaustive search we must take into account as key parameters the precision and the number of QoS parameters considered, for a given number of networks; these parameters influence strongly the computational effort.

We have considered a determined instance of the problem in order to validate the algorithm and perform the experiments. This instance is a scenario formed by two networks and up to four QoS parameters: throughput or bandwidth, delay, response time or latency, and cost (other QoS parameters can be easily added [_{0}, ..., w_{NQoS−1}} where NQoS is the number of QoS parameters considered in the experiments (2, 3 or 4) and the solutions satisfy the constraint given in

The experiments consider four possible profiles: profile #0 (all QoS parameters can have any weight from 0 to 1); profile #1 (all QoS parameters can have any weight from MINWEIGHT to MAXWEIGHT, where these variables have predefined values satisfying MINWEIGHT > 0 and MAXWEIGHT < 1); profile #2 (applications where the most important QoS parameters are delay and cost); and profile #3 (applications where the most important QoS parameter is the bandwidth). On the other hand, if NQoS = 2, we consider throughput (bandwidth) and cost; if NQoS = 3, we add delay; and if NQoS = 4, we add response time (latency). Finally, we have considered three possible precision degrees for h: 0.05, 0.01 and 0.005.

SEFI produces as output, for each profile and network, the following data: computing time, number of generated combinations, number of these combinations satisfying the restriction given in

We have used reconfigurable hardware technology in order to design and implement the SEFI digital architecture. Reconfiguration of circuitry at runtime to suit the application at hand has created a promising paradigm of computing that blurs traditional frontiers between software and hardware. This powerful computing paradigm, named reconfigurable computing [

We have used the ISE v13.3 technology and the FPGA devices provided by Xilinx [

Following with the last example, only 1,373,701 combinations are solutions. SEFI applies the fitness function to all these solutions in order to obtain the combination with the lowest fitness value; hence its corresponding network will be used for the VH decision.

We have also observed two interesting results, after examining detailed data from many other experiments. On the one hand, the best network could be any of the considered ones in any time, showing the importance of searching within a wide space of solutions. On the other hand, when the number of QoS considered parameters grows, other network different to the best one previously found can emerge now as the best.

Obtaining the best solution becomes slower to compute when considering more QoS parameters or more precision, because of the increased number of generated combinations.

Depending on the number of the discovered networks and their characteristics, the type and technology of the mobile sensor, the requirements of the application, and so on, SEFI can initially adjust the precision, profile and number of QoS parameters in order to obtain solutions in real time or in an acceptable time.

The aim of this paper was twofold: (1) to develop a fast algorithm to search all the possible combinations of quality-of-service weights in order to determine the best network for the vertical handoff decision performed by a mobile wireless sensor, given a determined heterogeneous scenario of wireless networks and the user’s preferences; and (2), to design and test an embedded processor with reconfigurable hardware technology to run the algorithm taking into account the constraints of the problem and the requirements needed in mobile sensor devices for dynamic environments: fast computation, small area, low power, and low economic cost. This was the first time that an algorithm for the vertical handoff decision phase based on QoS has been implemented in reconfigurable embedded processors. The results showed that the proposed architecture was able to achieve an acceptable performance.

Nevertheless, we have verified that the computing time increases a lot when we consider more networks, more quality-of-service parameters and smaller intervals searching the solutions, because the number of generated combinations becomes huge. This makes it necessary to enable mechanisms to select precisions that permit finding solutions in an acceptable time. We think this is a good starting point to add intelligent techniques to the algorithm in order to obtain good solutions in hard-computing scenarios without loss of precision or quality. This approach is our current research line.

This work was partially funded by the Spanish Ministry of Science and Innovation and ERDF (the European Regional Development Fund), under the contract TIN2008-06491-C04-04 (the MSTAR project), and by the Government of Extremadura, with the aid GR10025 to the group TIC015.

Heterogeneous wireless sensor network with three networks based on sensor routers of different technologies, giving different quality of service, and ready to provide support to mobile wireless sensors.

A mobile wireless sensor configures a path where several routers belonging to different heterogeneous networks can be reached in order to establish a link. This scheme is based on the scenario of

The complete scenario for the heterogeneous wireless sensor network where different processes for vertical and horizontal handoffs can be given.

SEFI searches the best combination of QoS weights for different networks.

Integration and behavior of SEFI in the architecture of a MS device.

Experimental results: Generated combinations for each profile and precision.

Experimental results: Combinations as solutions, for each profile and precision.

Experimental results: Computing times, for each profile, precision and FPGA.

Main characteristics of the implemented architecture.

Prototyping board | Digilent Nexys2-500 | Xilinx XUPV505-LX110T |

FPGA | Spartan 3E: xc3s500e-fg320-4 | Virtex5: xc5vlx110t-ff1136-1 |

Processor/memory | Microblaze with FPU and PLB/external | Microblaze with FPU and PLB/256KB local |

Operating system | Standalone | Standalone |

Clock frequency | 50 MHz | 125 MHz |

Occupied slices | 43% | 10% |

Power consumption | 0.097 W | 1.34 W |