Optimized Prophet Address Allocation (OPAA) for Body Area Networks

Transcription

1 Optimized Prophet Address Allocation (OPAA) for Body Area Networks Samaneh Movassaghi, Mehran Abolhasan and Justin Lipman Faculty of Engineering and Information Technology University of Technology, Sydney Sydney, NSW 2007, Australia s: {Seyedehsamaneh.Movassaghigilani, Abstract Each node in a Body Area Network (BAN) needs to be assigned with a free IP address before it may participate in any sort of communication. This paper evaluates the performance of an IP address allocation scheme, namely Prophet allocation to be used for BANs. This allocation scheme is a fully decentralized addressing scheme which is applicable to BANs as it provides low latency, low communication overhead and low complexity. Relative theoretical analysis and simulation experiments have also been conducted to demonstrate its benefits which also represent the reason for the choice of this allocation scheme. It also solves the issues related to network partition and merger efficiently. Keywords- address allocation; BANs I. INTRODUCTION (HEADING 1) In recent years, BANs have emerged as a promising technology, which will enable the interconnection of miniature, lightweight, low power monitoring devices through wireless communication links. The collection of these tiny sensor nodes which compose the BAN, aim to improve the accuracy and speed of the way data is recorded. For such a reason, various sensors and actuators are placed in different parts of the body for monitoring and logging data which is then transmitted and stored in a base station [1]. However, several design issues must be addressed in order to enable the deployment and adoption of BANs. At hardware level, body sensors must be small, thin non invasive, wireless enabled and must be able to operate at a very low power level [2]. The initial purpose of these networks was to interconnect information appliances attached to different parts of the human body. The standards for wireless communication in body area networks have been considered in IEEE [3]. In any network and consequently a body area network, each node has to be given a unique address. The devices which build up these nodes are either sensors, actuators or other processing and communication facilities. The nodes need to be capable of sending information to the health monitoring devices within an appropriate time in order to provide the medical staff with appropriate analysis of data. Consequently, an appropriate addressing scheme is required to achieve such an aim [4] Recently, various methods of addressing have been proposed for Personal Area Networks (PANs) using different technologies. These addressing schemes normally rely on IP addressing, mathematical graph algorithms or have been even based on some transmission schemes such as TDMA, FDMA, and CDMA. However, no known paper has ever come up with a method that solves the addressing issues in BANs or even wider networks based on BAN. In [5], an address planning scheme has been proposed for Zigbee based PANs using tree based arithmetic series. The method eliminates address duplication and wastage; however it is incapable of handling the issue of node mobility. An effective address allocation mechanism has been proposed in [6] which has the advantage of providing an efficient way of allocating a 16-bit address space without wastage and also provides the possibility of network expansion without any limitations. Additionally, it provides real-time addresses for all nodes and is seen as a good way of supporting node mobility. However, it is a centralized addressing scheme and results in significant communication overhead. Prophet address allocation scheme has been mathematically analyzed in [7]. Prophet scheme is a fully distributed scheme that enables the nodes to be addressed according to the state vector that has been provided by their parent nodes. In [8], Prophet addressing scheme has been used for large scale MANETs and it has been shown that by using a large address range, the address conflict problem can be avoided. It outperforms other addressing schemes proposed in MANETs in terms of latency (time needed to generate a unique IP address) and communication overhead (total number of control packets generated to ensure uniqueness of a new IP address). Based on these characteristics, it has better scalability compared to other MANET addressing schemes [8]. The idea in [8] has been developed in [9] to propose a method for secure addressing of MANETs. Under IP spoofing attack, each node can be re-addressed using the modified version of Prophet without adding any address interference. In fact, the authors in [8], add a random number to the state vector of each node to overcome the problem of address conflict. However, the idea given in [8] and [9] cannot be directly used for BANs, as they are multihop and will have collision if the original Prophet address allocation is used. In other words, in the original prophet address allocation via increasing the number of hops by introducing a second level, there is significant increase in the number of collision. Therefore, there needs to be some sort of modification to overcome this issue /11/$ IEEE 2098

2 In this paper, the Prophet address allocation scheme is studied and is considered to be a convenient addressing scheme for BANs due to its characteristics which will be described in detail as we go through. However, for it to be useful in BANs there is a need for some modification towards its address assignment. We have gone through this modification and proposed an addressing scheme called, Optimized Prophet Address Allocation (OPAA). Our main contribution towards optimization for the Prophet addressing scheme for BANs are as follows: Developed a real-time, small address space, low power transmission addressing strategy for BANs based on the Prophet address allocation scheme. Our proposed addressing scheme supports network expandability, provides efficient address space usage, and solves the issue of address duplication. Performed detailed simulations to measure the number of collisions. This paper is organized as follows. In Section 2, a description is provided on the system model. Prophet address allocation scheme is fully described in Section 3. The Optimized Prophet Address Allocation method has been discussed in Section 4. Finally, we describe our conclusion in Section 5. II. SYSTEM MODEL A. Node configuration in Body Area Networks We consider a BAN to have three different types of nodes which are the coordinator, routers and child nodes. The end device nodes have no child nodes. The coordinator node acts as a gateway to the outside world, another BAN, a trust centre or an access coordinator. The coordinator of the BAN is the Personal Digital Assistant (PDA), through which all other nodes communicate. As an example if a node is on the foot, data needs to pass through other intermediate nodes to reach the PDA. These intermediate nodes are the routers. In other words routers have a parent node, possess a child node and relay messages. The end node in the foot is the end device. An end node is a barren node in the network. Therefore, it is limited to performing the application which is embedded in it [5]. Two different topologies have been considered appropriate toward BANs applications which are namely the star topology and peer-to-peer topology. In the Star topology, the nodes are only capable of talking to the coordinator. This topology provides two types of transmission modes which are beacon mode and non-beacon mode transmissions. In a beacon mode transmission, the network coordinator controls the communication process. More specifically, it provides network association control and device synchronization. It transmits a periodic beacon for defining the beginning and ending of a super frame. The user is capable of defining the duty cycle of the system and the length of the beacon period between the certain limits which are defined by the standard [8]. In the non-beacon mode, the network node can use CSMA/CA if required to send data to the coordinator. The node has to power up and poll the coordinator in terms of receiving data from it. However, this type of communication has its own advantages and disadvantages. The advantage of this transmission mode is that the receiver node need not power-up regularly in order to receive the beacon [8]. As each node in the prophet allocation algorithm is assigned with an address via the information provided to its parent node, there is no need for it to be synchronized with the coordinator. Therefore, the non-beacon mode of transmission is considered to be convenient. However, the star topology does not work in some cases due to poor connectivity between the nodes on the ankle and the waist. In these cases a two hop star topology is of significant convenience which we have considered for our address allocation scheme. In our configuration, it is assumed that the coordinator is set to the first level, and that there are a number of routers in the next level. Child nodes are capable of joining the network at the third level. Also, each child node is assumed to know its parent node. B. Issues in addressing for Body Area Networks In terms of address assignment, there are some issues related to the configuration of BANs which affect the addressing scheme. One of which, is the short address space in BANs, which is basically reserved for the identification of the children, routers and network coordinators. As a matter of which, the number of child nodes and routers which can be supported will become considerably limited. One other issue which needs to be considered is the fact that the nodes in a Wide area Body Area Network (WBAN) are mobile and so may join and depart the network whenever they desire. As a matter of which, the allocation scheme will need to support it. In terms of network expandability, the network should be capable of expanding in size when required. Address interference, address wastage and duplication of addresses being assigned should also be taken into account. III. PROPHET ADDRESS ALLOCATION SCHEME The methodology in this algorithm is such that each node receives a black box which is shown in Fig.1. This black box allows the calculation of new addresses upon request. Each black box requires an initial node address a and a seed x 0. The function g is used to update the internal state x t. The address derivation function, f, uses a and x t to produce new addresses. m is the effective address range, in other words the bound on the number of nodes which can be assigned with an address with this allocation scheme. 1,,,,,

3 Figure 1.Prophet methodology The function h is the product of the first n prime numbers p i put to the power x i t. Initially, this matrix gets the value zero for all of its columns. g is a function which is used to update the internal state x, 0. Additionally, the initial value of a is a 0. An integer number is randomly chosen from within the range of [1, m-1] as the initial address, a 0, of the network. However, a 0 has to be coprime to m. It is also important to note that the use of different initial addresses or seeds will lead to different integer sequences which have the following characteristics: There is an extremely long interval between two occurrences of the same number. There is an extremely low probability on the occurrence of the same number more than once in a limited number of different sequences which are initiated with different seeds. The step by step procedure is given as follows: (1) The first node in the network chooses an initial address a 0 and the initial state vector (0, 0, 0, 0). (2) When a new node wishes to join the network, the previous node uses the function to create an address for the new joining node. As a matter of which, its state is also updated. (3) Now, this new node uses the address being generated by the previous node and sets itself with the same state matrix. (4) Consequently, this new joint node is made capable of assigning an address to one other node in the network. The above algorithm is illustrated via an example shown in Fig.2. Assume that each node is represented via its seed and state vector. Based on the number of levels in a network, a value can be chosen for k, which we have assumed to be 4. We could have chosen it to have any value more than 2, as the values of the other components in the vector do not change. As can be seen, in Fig.2 the first node has a 0 to be equal to 3 and based on the function, the address of the new node considering the initial state vector (0,0,0,0) will be (3+( ) mod 13+1) = 5. In this condition, the state vector of the node assigned with the address 5 will be (1, 0, 0, 0). Additionally, the state vector of the coordinator will be updated to be (1, 0, 0, 0). In the case where the coordinator wants to assign a new Figure 2. Prophet algorithm address for a new joining node, the address will be calculated to be (3+ ( ) mod 13+1) = 6. So the index of the state vector increases by 1 in each state and its tuple depends on the level at which the address is being assigned. IV. OPTIMIZED PROPHET ADDRESS ALLOCATION (OPAA) FOR BANS Our proposed addressing scheme is referred to as Optimized Prophet Address Allocation (OPAA). In this addressing scheme, there is a requirement for appropriate parameter optimization to minimize the number of collisions. There are several parameters in Prophet that must be considered for use towards its application in BANs. One of the most important parameters is the address range that has been indicated by m in step (1) of the procedure. We assume that the network consists of one coordinator, 7 routers and 40 end nodes (Fig.2). In fact, there is a trade-off between the number of collisions in the addressing scheme and address space usage. The larger the address space the lower the number of address collisions. As so, the parameter m has been optimized such that the numbers of collisions are minimized. One other parameter that should be considered is the prime numbers that appear in function h. In the original Prophet scheme, p i s are the first n prime numbers. As body area networks only consist of 1-hop and 2-hop nodes, only 2 prime numbers will be used in the h function which are 2 and 3. Due to the fact that these 2 prime numbers are close to each other, the original Prophet will have a relatively large number of collisions in BANs even in the case where the value of m is chosen to be large. In essence, Fig.3 shows the simulation results for a small BAN with one coordinator, 3 routers and 8 child nodes which use the original Prophet with m equal to As can be seen in Fig.4, a collision occurs in address allocation of the 4 th node. In this case, collisions have occurred due to the closeness of prime numbers irrespective of the address range. Therefore, we should select more appropriate numbers for BANs. Through a series of simulation results, it has been shown that the most appropriate prime numbers for BANs are 3 and 7. Fig.4. shows the simulation result for a BAN with a coordinator, 7 routers and 40 child nodes. It is also assume that m is equal to

4 In terms of address space, the value of m should be minimized. Through simulation results, the minimum value of m for BANs to have no collision is calculated to be greater than 400. In fact, each address can be shown with only 9 bits. Simulation results have shown the proposed parameters to be also appropriate for larger networks. However, in some other addressing schemes, like the method that has been presented in [10], there is a probability for address interference. Hence, there should be a method to rearrange the address conflict. But, via the enhancement towards the Prophet Allocation scheme, its parameters have been set in a way which reduces its collisions. On one hand, due to address space constraint and network size limitation in BANs, the proposed scheme should be capable of providing conflict free addresses to each node with reasonable transmission time whenever there is a request for it. Compared to other addressing schemes which have the issue of address collision, the proposed scheme is conflict free and there is no requirement for a transmission or mechanism to detect address collisions and solve the conflicts. A. Performance Comparison We have performed simulation of the proposed scheme in MATLAB (version R2009b) in a two hop star topology. Table 1 compares the original Prophet address allocation scheme with the modified Prophet scheme proposed in this paper in the sense of the percentage of collisions for different network sizes As can be seen, the original Prophet scheme has a large number of collisions even in cases where m is chosen to be a relatively large number. On the other hand, in the case that m is chosen to be a relatively large number in the proposed scheme there will be no collisions. By m=n-1, we mean the smallest prime number bigger than n. It is also important to note that by increasing the depth of the network, the performance of the original prophet scheme and proposed scheme in terms of the collisions will be decreased when we choose m to be close to the total number of nodes. In cases where m is relatively large, the original Prophet scheme is not capable of assigning collision-free addresses. This drawback can also be seen in Fig. 3 which shows two address conflicts in a small network where we choose m to be equal to In fact, multihop networks with more than 2 nodes in each level, will always have more than one address conflict. Figure 3. Performance of original prophet on small BAN Figure 4. Modified Prophet for large scale BAN TABLE I. collision percentage of original Prophet and proposed scheme N Network Structure Original Prophet OPAA Depth Routers Child nodes % 7.6% 15.3% 0% % 7.0% 19.3% 0% % 12.7% 24.0% 0% % 23.4% 26.0% 4.5% % 5.0% 25.1% 0% % 14.0% 30.0% 3.2% % 18.4% 34.3% 9.4% % 23.0% 39.0% 10.5% % 37.6% 41.2% 32.8% % 43.3% 44.3% 43.1% According to Table.1 and Fig.5, our method has shown a significant improvement in terms of address collision percentage compared to the original Prophet algorithm for up to 100 nodes. We have considered m to have a constant value of 1000 for the m parameter of the network for the usage of any number of nodes. The bigger the size of the network, the bigger a value has to chosen for the m parameter. In essence, if m is chosen to be 5000, in a network with N=364 nodes; we will have no collisions. In other words, our proposed scheme can support scalability on the condition that an appropriate value has been chosen for m. As an example, if m equals 1000 we can support a network of 100 nodes without any collision occurrence. In other words, the ratio of m/n has to be in the order of 10. On average, the percentage of address collisions in our addressing scheme if setting m to be N-1 is reduced by 10%. However, in cases where the number of nodes exceed 200 nodes the collision percentage is considered to be the same. m=n-1 m=1000 m=n-1 m=

5 in the network. In fact, each node need only communicate with its own parent node which is significantly advantageous when compared with centralized addressing schemes such as the LAA algorithm proposed in [7]. Figure 5. Collision percentage versus the number of nodes and choice of m Based on what we had expected through theoretical analysis, the performance of both allocation methods has improved by only changing m to a bigger value which is set to In fact, it has had a major decrease of approximately 20 % in the ratio of collisions by increasing the value of m to It is also important to note that Fig.5 shows the numbers in the table 1 graphically. This has caused a few jumps in the figure, which is related to the increase in depth of the network. In essence, we have 2 levels for 111 nodes and 4 levels for 121 nodes. Our allocation scheme which is based on increasing the value of m and choosing big steps of 3 and 7, as the prime numbers in our state vector, has no collision if m is set to However, the original Prophet address allocation scheme has 5-10 % collision for the same value of m. Given an increase in the number of nodes, the number of collisions in our allocation scheme has a slight increase which however is much less (around 10%) compared to the original prophet allocation scheme. Fig.5 has shown the same amount of collision percentage for a larger number of nodes, N. For example, for N=781, the collision percentage of all methods is about 45%. In fact, the address collision performances of all methods are the same when the values for m and N are chosen close together and relatively large. Due to the limitation on the computation power and memory capacity of mobile nodes, the addressing scheme for BANs should be as simple as possible. As can be seen in table 1, our allocation scheme, importantly, has no collisions for a network with a depth of 2 with a number of 100 nodes. Additionally, our proposed scheme adds no complexity to the original prophet scheme and consequently has many advantages over other methods as well as the original Prophet with respect to computational complexity. Finally, the Prophet address allocation requires no center node such as a coordinator to allocate addresses to other nodes V. CONCLUSIONS Our approach considers the usage of the Prophet address allocation scheme for Body Area Networks. It is found to be convenient in terms of achievement of low communication overhead, low complexity, low latency and high scalability. Enhancements to the Prophet addressing scheme have been proposed which reduce routing traffic and address collisions. As most application of BANs require less than 100 nodes and simulation results have shown a slight percentage of collision via this number of nodes, our method is considered to be well suited for application in BANs. However, with a little more effort, the allocation scheme could be improved such that it supports any BAN with any network size while observing key factors of latency and communication overhead, this research forms part of our future work. If an appropriate value is chosen for m, the number of collisions will have a major reduction. m can also be set as a dynamic value based on the size and depth of the network which is considered as our future work. REFERENCES [1] Chen,M. Gonzalez, S. Vasilakos, A. Cao, H. and Leung,V.C.M Body Area Networks: A Survey 2010, Mobile Communication and Networks, Springer, 2010, DOI:1007/s [2] M. Quwaider, and S. Biswas On-body Packet Routing Algorithms for Body Sensor Networks, First International Conference on Networks and Communications. NETCOM '09. [3] IEEE Standard for Information Technology Part 15.1: Wireless medium access control (MAC) and physical layer (PHY) specifications for wireless personal area networks (WPANs), IEEE Standard Working Group Std., [4] J. Espina, T. Falck, and O. Mlhens, Network topologies, communication protocols, and standards, Body Sensor Networks, Ed. Javier Espina, Springer, London, December [5] J.I. Agbinya, M.A. Agbinya, "A New Address Allocation Scheme and Planning of Personal Area and Sensor Networks," International Symposium on Parallel and Distributed Processing with Applications, ISPA '08., Proceeding, pp , Dec [6] H.I.Jeon and Y.Kim;, "Efficient, Real-Time Short Address Allocations for USN Devices Using LAA (Last Address Assigned) Algorithm," The 9th International Conference on Advanced Communication Technology, vol.1, no., pp , Feb [7] C. Lauradoux, and M. Minier, A Mathematical Analysis of Prophet Dynamic Address Allocation, inria , version 1-4 Nov [8] H. Zhou, L.M. Ni, M.W. Mutka, "Prophet address allocation for large scale MANETs," INFOCOM Twenty-Second Annual Joint Conference of the IEEE Computer and Communications. IEEE Societies, vol.2, pp vol.2, 30 March-3 April [9] Hongbo Zhou, Secure Prophet Address Allocation for Mobile Ad Hoc Networks, IFIP International Conference on Network and Parallel Computing, NPC Proceeding, pp , Oc t [10] C.E. Perkins, J.T. Malinen, R. Wakikawa, E.M. Royer, Y. Sun, IP address autoconfiguration for ad hoc networks, July 2002, Internet DRAFT, DRAFT-ietf-manet-auto-conf-01.txt. 2102