Determining Fused Information Priority and Organizing Information Flow - Assignment Example

Data Fusion and Delay Tolerant Networks Determining fused information priority and organizing information flow 1. Introduction RFID or Radio Frequency Identification is becoming more popular due to lower tag cost, sizes, and increasing communication range. Similarly, DTN or Delay Tolerant Network provide data transmission opportunities for networks with intermittent connectivity such sensors composed of RFID tags and readers in a hostile environment where continuous end-to-end connection is not possible. The following sections discuss the possibility of data fusion in delay tolerant networks and establishing fused information transmission priority and efficient information flow in networks with sporadic connectivity. 2. Data Fusion Strategies and Delay Tolerant Networks Data Fusion Strategies Data fusion according to Klein (2004) involves numerous processes such as detection, association, correlation, and estimation in order to combine data and information from one or more sources (p.2). Data fusion in multi-sensor system therefore is incorporation of different data about events or objects being observed gathered from one or more sensors. Multi-sensor fusion actually deals with data integration or the fusion of sensory data from multiple sensors (Joshi & Sanderson, 1999, p.18) Multi-sensor data fusion can be achieved in different ways depending on sensor configuration (complementary, competitive, and cooperative) and fusion type (fusion across sensors, attributes, domains, and time). A sensor configuration is complementary when sensors are not directly dependent on each other but can be fused in order to get a full picture of phenomenon observed (Mitchell, 2007, p.6). A competitive sensor configuration on the other hand is totally independent with each sensor delivering their own measure of the same property (ibid, p.6). In cooperative configuration, each sensors “cooperate” and contribute information that would not be available from single sensors (ibid, p.6). Some strategies consider the first type of fusion (fusion across sensors) doing the same functions such as measuring temperature and so on while others concentrate on attributes of sensors measuring different quantities such as air temperature, pressure, and humidity. Others use attributes (fusion across attributes) such as air temperature, pressure, and humidity while some fuse data across domains using the same attribute mentioned above. Fusion across time is when current sensor measurements are fused with historical information in order to enhance system accuracy (Mitchell, 2007, p.5). Other strategies rely on sensor configuration such as combining independent sensors in order to generate a more understandable image of the phenomenon under observation or use information provided by two or more sensors in a cooperative configuration to come up with specific information not available from a single sensor (ibid, p.6). Cooperative sensing as mentioned earlier, nodes performs measurements independently and makes soft or hard decisions. These binary decisions are then forwarded to a cognitive base station or CBS where these decisions are fused before making the final decision to transmit (Rahim, 2009, p.129). Most data fusion methods used in retrieving information is based on score methods where relevant scores are required to determine the differences in retrieved documents. Score method or score normalization (linear and non-linear) in data fusion gather relevant scores from all retrieved documents then normalized them for fusion. For instance, linear score normalization method normalize scores based on their range as illustrated by the function below. Figure 1- Linear Method (Wu, 2012) However, according to Wu (2012) such approach is not consistent as some information retrieval systems only provide a ranked list of documents rather than scores. Moreover, these scores if ever produced are generated in different manner thus; normalization of scores is required before any comparison can be made (p.19). Note that this normalization of scores generated by each node in a wireless network and subsequent comparison occurring in a retrieval system facilitate more accurate data fusion. For instance, in a wireless sensor network including those using RFIDs, nodes transmit the locally processed data to a fusion centre where information is combined or fused to produce specific information (Xia, 2001, p.3). This approach can be enhanced by scoring and comparison to ensure that fused information is relevant. For instance, DFuse is a data fusion framework mainly intended for video streaming applications that include assigning aggregation roles the nodes within the network. In this approach, decision is made using cost (transmission cost, power variance, and ratio of cost to power), as condition in determining the suitability of a node in assuming a specific role (Misra et al, 2009, p.195). This is approach is similar to scoring but using cost of transmission and other constraints to qualify each information for fusion. Data fusion strategies often support physical data fusion where various decisions are fused in a fusion centre in order to enhance the final decision where and when to send the information as discussed earlier with wireless sensors. It also supports data fusion frameworks in order to obtain further knowledge of networked environment (Shahbazian et al., 2005, p.503). Delay Tolerant Networks Some networked environments operate intermittently or delay tolerant networks where traditional routing protocols may find it difficult to achieve adequate performance. For this reason, new routing approaches were introduced to address issues involving frequent, arbitrarily, long-lived connectivity disruptions. These include deterministic or scheduled, enforced, and opportunistic routing approaches. Using a modified version of the Dijkestra algorithm, which is based on scheduled contacts, deterministic routing can route messages from a source to destination in case of disruptions (Vasilakos et al, 2012, p.6). To prevent packet dropping, enforced routing ensures that resources are allocated based on packet priority and forward information based on effective packets and routing performance (Li et al, 2009, p.1). Opportunistic routing on the other hand, decide whether to forward a message or carry it further over a network based on message vectors as nodes encounter other nodes (Vasilakos et al, 2012, p.6). Data Fusion and determining information priority Primarily, data fusion involves low and high level processing. Low-level processing deals with target detection, classification, identification, and tracking while high level processing involves situation and threat assessment (Klein, 2004, p.95). Low-level processing requires algorithms based on physical models, feature-based inference, and cognition while data fusion involving assessments deals with current relations among objects and events found within the operational environment (ibid p.96). Studies concerning data fusion in a multi-sensor environment include Girija et al, (2000) “Tracking Filter” that integrates information from multiple sensors in order to produce a specific unified data about entity, activity, or event. Basically, MSDF’s tracking filter involves linear and nonlinear estimation techniques such as the U-D factorized Kalman filter, state vector fusion filter, and fusion policies (p.159). An “Integrated Test Range” approach is also used by Bhattacharya and Raj (2004) as a multi-sensor data fusion strategy by classifying different of sensors such as fusion based on similar type of sensors, fusion based on dissimilar type of sensors, and priority selection based on the most accurate track data and reliability of the sensor. In this strategy, the fusion algorithm generates three different fused positions as mentioned above and output a single valid sensor data (p.239). An entirely different approach adopted by Bloch (1996) use information “Combination Operators” for data fusion rather than real values. These numerical fusion operators are classified from their possible behaviours that include context independent constant behaviour or CICB operators, context independent variable behaviour or CIVB operators, and context dependent or CD operators. The first operator or CICB can found in other mathematical frameworks such as the probability and Bayesian fusion, fuzzy sets and possibility theory. Here the degree of belief is measured by probabilities (a priori, conditional, and posterior probabilities). The CIVB operators on the other hand use the symmetrical sums from fuzzy sets and possibility theory as operators. These are hybrid aggregation operators possessing either conjunctive or disjunctive, or compromising behaviour. Context dependent operators or CD operators take into account selected contextual information about the sources such as conflict between the sources and reliability of the sources (p.56). The SMART fusion strategy or Space and airborne Mined Area Reduction Tools developed by Bloch et al. (2007), is highly reliant on Dempster-Shafer evidence theory or DS, which is widely used in satellite image process (p.176). The fusion strategy considers each class (with confidence values) as one information source. When no confidence value is present, the information is ignored. Another is adding a global discounting factor or BF1 where classifier is again considered as one information source but the capacity to fuse depends on the singletons and D or “discounting” (see equation below). m (D) = 1 - α. Contrary to “confidence value” strategy, this fusion approach takes into account that some classes may be undetected thus its values should be considered equal to 1 and 0 for detected classes. The “discounting” equation above us a discounting factor α equal to the sum of the diagonal elements of the confusion matrix, divided by the cardinality of the training areas. Other fusion strategies are introduced with SMART including analysis behaviour of each classifier for each class and use of confusion matrices for more specific discounting (Bloch et al, 2007, p.178). To enable application of Intelligent Transportation System or ITS, Geisler et al, (2011) developed an architecture that would detect and provide automatic priority-based processing and validation of event messages for subsequent data fusion. In this approach, mobile phone data are sent to a database server that will clean, transform, and fused the data. Note that this strategy is for mobile phone but can be implemented using RFID as they can both transmit and fused data. The most essential component in terms of prioritization and fusion is the “Event Validator” tasked to process, verify, and prioritize messages. First, it verifies an event using information from the data stream then calculates the quality or the level of reliability of the event (p.4). For instance, vehicles create messages and these messages containing information such as current position of the vehicle, acceleration, braking, and other information specific to a certain type of event are being sent to the Reflector Traffic Information or RTI server for processing. This strategy is made possible by multiple positioning techniques in cellular network that can detect the location of a certain sensor, which can be a mobile phone or a long-range RFID. The idea is to derive traffic parameters for a particular road segment using signals from different sensors (p.7). However, since accuracy of mobile phone (this can be an RFID sending information) location as well as identification of the road segment varies from one cellular network to another, some form of monitoring system that can facilitate a good estimation of the required information. For instance, the accuracy of this technique may be affected when getting the location of a bus containing 20 travellers all carrying a mobile phone. WLT-based or Wireless Location-Based monitoring system is one good candidate as it determines the quality of the estimated data by comparing the current data from the true data gathered from actual road traffic. Cleaned data (those that were received and checked) are sent to the aggregation nodes where they are processed base on their time stamps and road segments. Here the data are rated and prioritized using established criterion. Priority in this approach is determined by the urgency of the information (p.8). Organizing Information Flow in DTN The architecture of Delay Tolerant Networks makes it compatible to a number of operating environments including routing in mobile ad-hoc networks or MANET and connecting any pair of nodes (Vasilakos et al, 2012, p.5). In other words, DTN can conveniently forward, store and carry forward information even in different networks. However, the most useful component in terms of data fusion and organization of information flow is its custody transfer service where information can be stored in a certain amount of time until it is safe to transfer. For instance, Fall et al, (2011) investigation of custody transfer mechanism suggest that custody transfer in DTN improve the reliability of message forwarding activities by creating “custodian” nodes that will hold the message until another node is free to accept the message (p.1). Integrating MANET and DTN can have a number of benefits and in particular enable MANET applications to make sensible choices when interacting with DTN. For instance, result of the investigation conducted by Ott & Dwertmann (2006) suggest that MANET routing protocols running in a DTN performed well particularly in making message forwarding decisions (p.221). Passive RFID tags and readers in a distributed sensor system may perform information gathering, data acquisition, delivery, and storage particularly when supported by a special DTN with unique communication and storage constraints. The Featherlight Information Network with Delay-Endurable RFID support or FINDERS developed by Yang and Wu (2009) can be use to expand use of RFID in wireless network aiming to aggregate or fuse information. This strategy is consisting of two types of nodes –RFID readers and tags deployed in strategic locations as shown below. Figure 2 - FINDERS configuration (Yang & Wu, 2009) FINDERS is designed to detect events of interest and fused information through intermittent wireless asymmetric communication between passive RFID tags and readers. FINDERS supports delay in data delivery and enable efficient data transmissions by dynamically creating redundancy, managing data queues, and routing data packets. An opportunistic network, FINDERS employs EDC or Effective Delivery Capability of tags and readers using time synchronization. This new data transmission protocol consists of different phases that include data retrieval where each reader periodically broadcast query signals until it detects nearby tags. The reader then read detected tags data and EDC parameters and updates them according to calculated EDCs. It then performs data queuing where the reader inserts the data into its queue, which is sorted according to Fault Tolerance Degree or FTD. One or more nearby receiving tags will be identified for data transportation, which are usually the data packets at the top of the queue. Two or more copies of data are created and processed for each packet before data transmission (read and writing operation) to each selected tag (p.7). Result of simulation suggests that FINDERS performed well as it network throughput increases while the average packet delay decreased as the number of tags increased. 3. Conclusion Data fusion facilitate exploitation of data measured by different sensors and applications aiming to obtain a more comprehensive information about the system being analysed by combining sub-information from multiple sensors. Cooperative sensing may applied to an RFID system as tags and readers may be allowed to contribute information for data fusion. Similarly, fusion across attributes can enhance accuracy of decisions for data fusion particularly when a data fusion framework similar to DFuse supplements it. Opportunistic routing is also valuable particularly in making decisions about message forwarding and prioritization while RFID data fusion in special and opportunistic DTN similar to FINDERS can improve network throughput and packet transmission. More importantly, combination of cooperative sensing, fusion across attributes, and opportunities routing can facilitate priority-based data fusion and organized flow of information in a DTN network. 4. Bibliography Bass T, (2007), Intrusion Detection Systems & Multi-sensor Data Fusion: Creating Cyberspace Situational Awareness, Communication of the ACM, pp.100-105 Bhattacharya S. & Raj R, (2004), Performance evaluation of multi-sensor data fusion technique for test range application, Sadhana, Vol. 29, Part 2, pp. 237-247 Bloch I, (1996), Information Combination Operators for Data Fusion: A Comparative Review with Classification, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 26, No. 1, pp. 52-67 Bloch I, Milisavljevic N, & Acheroy M, (2007), Muti-sensor Data Fusion for Spaceborne and Airborne Reduction of Mine Suspected Areas, International Journal of Advanced Robotic Systems, Vol. 4, No. 2, pp.173-186 Fall K, Hong W, & Madden S, (2002), Custody Transfer for Reliable Delivery in Delay Tolerant Networks, Intel Research, Berkeley California, pp. 1-6 Geisler S, Quix C, Gehlen G, & Jodlauk G, (2011), A Quality and Priority-Based Traffic Information Fusion Architecture, Information Systems, RWTH Aachen University, pp.1-12 Girija G, Raol J, Raj R, & Kashyap S, (2000), Tracking Filter and Multi-Sensor Data Fusion, Sadhana, Vol. 25, Part 2, pp. 159-167 Joshi R & Sanderson A, (1999), Multi-sensor Fusion: A Minimal Representation Network, Word Scientific Press, US Klein L, (2004), Sensor and Data Fusion: A Tool for Information Assessment and Decision Making, SPIE Press, US Misra S, Woungang I, & Misra S.C, (2009), Guide to Wireless Sensor Networks, Springer, Germany Mitchell H, (2007), Multi-Sensor Data Fusion: An Introduction, Springer, Germany Ott J. & Dwertmann C, & Kutscher D, (2006), Integrating DTN and MANET Routing, SIGCOMM’06 Workshops 2006, pp. 221-228 Qinghua L, Zhu S, & Cao G, (2009), Routing in Socially Selfish Delay Tolerant Networks, National Science Foundation, pp.1-9 Rahim A, (2009), Interference Mitigation Techniques to Support Coexistence of Ultra-Wide Band Systems, Jorg Vogt Verlag, Germany Vasilakos A, Zhang Y, & Spyropoulos T, (2012), Delay Tolerant Networks: Protocols and Applications, CRC Press, US Wu S, (2012), Data Fusion in Information Retrieval, Springer, Germany Yang Z. & Wu H, (2009), Featherlight Information Network with Delay-Endurable RFID Support (FINDERS), National Science Foundation, IEEE 2009, pp. 1-9 Xiao J, (2006), Distributed Data Fusion and Information Processing in Wireless Sensor Networks, ProQuest, US Read More