Implementation of New Fault Tolerance Solution in Wireless Sensor Networks in a Multi-channel Context
Chapter One
Objectives of the Study
The main objective of this thesis is to provide an adequate solution to fault tolerance in wireless sensor networks in a multi-channel context. The general objectives are:
- To carry out state of the art of multi-channel communication and fault tolerance in wireless sensornetworks
- To implement a proposed algorithm for decentralized fault tolerance detection in wireless sensor
- To Simulate the implemented algorithm
- To compare implemented algorithm with already existing algorithm.
CHAPTER TWO
LITERATURE REVIEW
STATE OF THE ART
State of Art for multi-channel communication in WSNs
The main goal of multi-channel communication protocols in WSNs is to maximize energy efficiency in order to increase the network lifetime. Due to this advantage, the state of the art research, has proposed a good number of multichannel protocol for communication in WSNs. The authors in [8], proposed a tree-based multichannel protocol which enable data collection application thereby allocating channels to the disjoint trees in order to exploit parallel transmission of data among the trees. The application specific end to end communication delay which is critical to real time communication has not been really pay attention to in the existing work. In [29], a Multichannel protocol was proposed MCRT for such real time communication using a flow based channel allocation strategy, that allocate channels to network based on many to one data flows and this protocol includes a real time packet forwarding strategy.
Channel assignment with minimum interference known as 2-hop colouring problem allows repetition of colours, so in [2], proposes a Dynamic channel allocation (DCA) algorithm as a novel solution for the 2-hop colouring problem that optimally assign minimum channel in a distributed manner to avoid interference. In reference [30], A MMAC is proposed for ATIM window such that the multi channel phase is splited at beginning of the beacon interval. The nodes that have packets to transmit negotiate channels with the destination nodes during this window and if the channel is acquired the transfer can be done.
Ramakrishnan and Ranjan [31], proposed SMC MAC, which uses a single dedicated control channel and eight data channels. It initialized all the nodes to stay in control channels such that channel negotiation is done via RTS/CTS control packets and to transmit data the first free channel is selected. Also in [32], CAM MAC was presented that uses handshake which consist of three phase for communication and this protocol solves multi-channel problems using a single transceiver and completely eliminates synchronization through the exploitation of cooperation.
Discussion of Fault Tolerance in Wireless Sensor Network
The fault tolerance hierarchy is given based on different kind of fault in WSNs [13], this fault includes: connectivity fault, link fault, node fault and malfunctioning fault. The link fault needs relay with high power capabilities to be place between two nodes for sending information to it destination. Communication graph is used for identifying the fault.
Fault Types
There are a lot of authors who has presented fault types in their different opinion and this authors gives different fault types [12], [26], [28], although
there are similarities in their differenciation. The two main types of fault in wireless sensor networks deployment include node fault and transmission fault. The figure shows the fault types.
Figure 2.1: fault types
Fault tolerance occurrence at differentlevel
fault tolerance occurs at different level in wireless sensor networks. There are five level at which fault occurred in WSNs [24] such as Physical layer, hardware layer, system software layer, middle ware layer and application layer. According to the study, there are four main layer at which fault occurred [4], [22], [23] which include hardware layer, software layer, application layer and network communication layer.
CHAPTER THREE
IMPROVED DISTRIBUTED FAULT TOLERANCEALGORITHM
This chapter explicitly explained the improved distributed fault tolerance algorithm in wireless sensor network in a multichannel context. The bases of the improved distributed fault tolerance algorithm(DFD) is extracted from paper [17].
Network Model
We consider a multi-hop wireless sensor network deployment, where each node communicate with the help of a neighboring node. Then all sensor nodes are deployed geographically in an intended area. All the sensor nodes can have the same transmission range and can communicate to two hop nodes through it one hop neighbor. That means it can only communicate with it two hop nodes through it one hop node that is detected to been fault free. The sensor node deployment is shown in figure 1 with whole area covered with sensor nodes. The circles that is black is said to be a faulty (FT) node, the white circles are said be good (GD) node and the irregular shape inside the sensor deployment area is said to be a failed area in the deployment meaning no sensor nodes in the network can communicate with the sensor nodes in this area. We are using voting system but not dependent on majority voting, we assume that each sensor in the network area has atleast three neighbors. Because both large and small amount of sensor nodes can be deployed into interested area to form a wireless sensor network. But each of the node in sensor deployment can locate it neighbors within its transmission range through ACK protocol or sending a HELLO message. Considering multi-hop communication is to enable sensor nodes to be able to communicate to the fusion center or Cluster head in case of node failure in a multi-channel communication wireless sensor network deployment.
For example, probability of sensor failure [37]. Assuming p represent the probability of sensor failure and r denote the probability that a faulty node has a communication unit which is said to be fault free. Then if total number of nodes under detection is n, np nodes are faulty. N(1-r) nodes are unable to communicate with their neighbors. Just n(1-p) + npr nodes are involved in fault detection.
So the new probability is defined by:
CHAPTER FOUR
Simulation or Proofs of the Proposed Algorithm
This chapter explicitly stated the theorem of the improved proposed algorithm with proofs and example to show that the distributed algortihm proposed for multichannel communication is correct and can be applied to real life scenario where wireless sensor network is deployed. It also explained the simulation tool that was used in carrying out the implemented algorithm.
Chapter Five
RECOMMENDATION ANDCONCLUSION
Resolved Issue
In our improved distributed algorithm we were able to solve the problem of nodes getting wrong detection about it self which in turns leads to wrong detection status about other nodes in it neighbors. We achieved this be enforcing in our proposed improved algorithm by first getting the detection status of nodes with larger neighbors before the ones with fewer neighbors until the detection status of the whole nodes in the network is achieved. We also solve the problem of multi- hop and multi-channel communication, in this we stated clearly in our algorithm that any node whose detection status is good (GD) and has no good neighbor (I.e all neighbors faulty (FT)) is itself faulty (FT). Since it can not communicate it sensed data to it two hops count neighbor and it has no alternative channel to use in passing the details of it sensed data. Finally we were able to state each theorem of our proposed improved algorithm and we proved it.
Unresolved Issues
The simulation result was not fully achieved because Castalia 3.2 which was suppose to be our bases for simulation in the Omnetpp4.2 environment was unable to be imported to the platform of the Omnetpp4.2. This was the reason for proving the algorithm and part of our algorithm was some what implemented in Omnetpp4.2 with out the Castalia
Conclusion
In a faulty sensor node distributed detection algorithm where each node detect it own status to be either good or faulty and this claim will be supported by its neighbors because they also check the behaviour of the node themselves The probability of faulty sensor nodes being diagnosed as good and good nodes not been diagnosed as good is very low. The proposed algorithm ensure that sensor nodes with larger neighbors are been diagnosed first before the one with fewer neighbors.
Challenges
In our improved algorithm the main problem we faced is in the aspect of simulating the algorithm using omnetpp with castalia. Even though we were able to successfully install omnetpp4.2 and castalia 3.2 but we were not able to import castalia in omnetpp as described in the manual. So we were just able to deploy few sensors in the network.
Future work / Extension of Work
In our future work we want to have a real testbed where our proposed improved algorithm will be tested in real life scenario. We also wish to calculate the detection accuracy for each nodes in wireless sensor network. The detection accuracy is the ratio of the number of faulty sensors detected to the total number of faulty sensors in the whole network.
Reference
- Hill, R. Szewczyk, A, Woo, S. Hollar, D. Culler, and K. Pister, System Architecture Directions for Networked Sensors, ASPLOS, November2000.
- R. Chowdhury, N. Nandiraju, P. Chanda, D. P. Agrawal and Q. A. Zing, “Channel allocation and medium access control for wireless sensor networks”, Ad Hoc Networks, vol. 7, pp.307 -321,2009.
- Saleh, M. Eltoweissy, A. Agbaria and H. El-Sayed, “Fault Tolerance Management Framework for Wireless Sensor Networks”, Journal of Communications, Vol. 2, No.4, June 2007, pp. 38-48
- Chouikhi, I. El Korbi, Y. Ghamri-Doudane and L. A. Saidane, “A Survey on Fault Tolerance in Small and Large Scale Wireless Sensor Networks” Computer Communications 69 (2015) 22-37
