Home

Research Programme 3 (RP3)

RP3 - Autonomic Sensor Communites
Programme Leaders: Prof. Gregory O'Hare, Prof. Paddy Nixon, Prof. Noel O'Connor and Dr. Cian O'Mathuna

Context and Objectives

The paradigm of widespread sensing that underpins the CLARITY research vision will yield data of an unprecedented volume and heterogeneity. Very often we think of sensing as a relatively straightforward process of data capture given an appropriate sensing infrastructure. However, this ’blind’ gathering of data is an overly simplistic view, which naively fails to consider the use to which the data will be put, and the power envelope within which it must be assembled. In this RP we will address these shortcomings. In practice, this means that not only should this research programme provide mechanisms for harmonising and delivering sensed data to RP4, but it should also ensure that these same mechanisms can be moderated based on personalization and retrieval feedback (RP5 & RP6). This RP addresses several key research challenges. It will develop a middleware framework which will facilitate the management of autonomic wireless sensor networks. It has been recognised that such architectures must exhibit a narrow waist [1] offering a translucent Sensor Protocol that exports neighbour management and message pools [2].It will harness the potential of agency to distribute decision making across the various layers of the system and network partitions, to facilitate robust collaborative behaviours adapting and reflecting evolving resource constraints. It will develop novel techniques for adaptive capture and filtering, data aggregation and dissemination within a dynamic, unreliable, unstructured, decentralised and resource constrained network.

Within CLARITY this RP will play a central role propagating aggregated, harmonised, and filtered data to http://www.clarity-centre.org/content/research-programme-5-rp5 Content Processing for Extracting Context & Semantics. RP4, RP5 and RP6 will influence data capture, filtering and aggregation (WP 3.1 and 3.2). It will also inform adaptivity at the platform (RP2) or materials (RP1) level through energy scavenging or conservation, while providing basic software building blocks for the demonstrators.

Work Packages

WP 3.1: Adaptive Capture and Filtering

Effective decision making within WSNs is often hindered by the absence of appropriate architectures with which to aggregate sensor data and achieve effective filtering techniques. The design of sensor aggregation architectures to accommodate and aggregate a myriad of sensor types, delivering different media data simultaneously, is the target of this workpackage. Individual sensors or sensor subnets are capable of capturing multi-modal input and enhancing or degrading this data portfolio in the light of some operating context. This context would not only consider the power envelope of the sensing devices per se, but also the perceived criticality of those events sensed to specific application needs. This workpackage will address the need to filter a potentially large and diverse set of heterogenous sensor data enabling salient feature extraction (WP4.1) and appropriate actions can then be invoked (RP5 & RP6). Such actions could manifest itself in the personalization layer as feedback to end users, e.g. content adaptation. Alternatively feedback could be propogated downward to the platform and sensor network level resulting in network topology adaptation, either at the individual sensor level or collectively deriving from negotiated sensor cluster behaviour. Key research tasks involve developing filtering models together with application and power sensitive aggregation algorithms based on semistructured data approaches.

WP 3.2: Sensor Aggregation

This WP explores techniques for the fusing of sensory data. Typically this will necessitate shared data anchors which assist in performing key sensor data operations. Examples might be aggregation based upon data or sensor anchors, defuzzification to reduce uncertainty, enhancement to extend or improve data content, aggregation based upon multi-media sensory sources or filtering of content based upon noise. Key WP research tasks involve the identification of a sensor ontology and a naming convention for individual sensors and sensor clusters. The identification of key data sensor anchors (such as location, time, calibration/quality etc.) can be used to automatically tag data sets facilitating aggregation. It will also derive algorithms for aggregating complex multi-modal sensor inputs from heterogeneous sensors of varying quality and sophistication;

Numerous approaches have been adopted in the literature including Continuous Query (CQ) [3], Directed Diffusion [4] and concave non-decreasing cost functions [5]. Many approaches have been predicated by the naive assumption that an intermediate node can simply aggregate incoming packets into a single outgoing packet [6].Such algorithms are complex not only because of the inherent data diversity but also because of the scale of the networks of sensors and the need to provide dynamic changes in the algorithms to accommodate specific application queries. This demands autonomic aggregation and capture framework.

WP 3.3: Delivery and Dissemination within an Autonomic Infrastructure

Reliable and scalable information dissemination is critical for sensor networks. Centralised solutions are widely adopted because they promote transparency and uniformity, however the centralised paradigm does not suit highly dynamic systems as typified by Ad hoc networks and MANETs failing to efficiently accommodate unstructured networks where nodes join and leave intermittently, where the delivery of some parts of data sets is not reliable and where network partitions can appear.

In this work package we attack the problem of scalable and reliable data and context dissemination. We draw heavily upon the personalization and feature extraction work undertaken in the Personalization in Context and Contextual Content Analysis research programmes and use these to drive the development of new gossiping, or epidemic, messaging models. We will attack the problem in two ways. Firstly, rather than using widely adopted flooding models, we will develop self-managing (autonomic) systems models. By considering the query as a key driver in the system we investigate epidemic style algorithms to intelligently place data and propagate data through the system. This is intimately related to workpackages 3.1 and 3.2. Secondly we will consider the persistence of data and the notion of update propagation and the impact of reliability models on the epidemic dissemination models. Data propagation (both network and application) will adhere to the general principle of as and when needed. This will necessitate intelligence in the determination of need. Two key metrics underpin our approach those of query robustness and data stability.

WP 3.4: Collaboration, Autonomy and Mobility

It has been recognised that many decisions taken within WSNs could be improved if made collaboratively. Necessarily collaboration demands communication and which in turn results in power burn. Agency plays a key role in supporting collaborative and distributed decision making. Specifically this work package investigates mechanisms for supporting decisions that are inherently collaborative in nature. The apparatus developed here will be used in the delivery of intelligent decision-making within the other work packages. Agency advocates a paradigm where stakeholders collaborate and negotiate in the solution of shared problems. Agents can assist in the decomposition and management of the decision-making process. As the nature of decisions change, depending upon context, decision-making is performed within certain constraints, time, resources, quality and these influence the process and associated outputs. Recent research has explored agents as a delivery vehicle for imbuing WSNs with autonomic capabilities [7][8].

If we consider the system for the capture, analysis and ultimate delivery and presentation of sensed information as both layered and partitioned, then agents could provide a novel and intuitive way of facilitating both cross-layer and cross-partition decision-making. Furthermore, they facilitate managed knowledge sharing across the various layers and/or partitions. For example an agent could encapsulate a query on a sensor network which could be used to both enable the delivery of the query and to manage the underlying network resource structure. Such an agent would necessarily adapt its behaviour by, for example, varying network transports and partition policies whilst also changing the behaviour of its collaborating resources. This all has to be achieved in a stable way balancing unconstrained learning models with policy based management.

WP 3.5: Utility Based Autonomy

Adaptive, sensor-rich systems exploit rich combinations of heterogeneous distributed sensing nodes to a range of other problems from resource conflicts to privacy and trust conflicts. Trustworthiness is the over-riding metric by which all such conflicts can be measured and managed. Trust, therefore, influences the amount of information that a reasoner is inclined to reveal, while utility analysis allows us to evaluate the expected benefit that would motivate agent collaboration. We consider trust to embrace such social traits as dependence, confidence and collaboration. We consider trust, as a core enabler, in balancing the complex trade-offs demanded by adaptive sensor rich systems, systems where the decisions of individual nodes have an impact on the well being of the overall network. Game theoretic approaches have been used to model agent disposal to collaboration while issues of trust in agent-based systems are now being addressed [9][10].

Decisions pertaining to information exchange, encryption, access control, migration and resource sharing, intelligent power management, coverage and routing all imply system choice. Trust-based infrastructures will provide the mechanisms through which systems can ground such decision processes informing their perspective of the risks and benefits involved. In this work package we attack this problem from two perspectives. A strong relationship exists between context and utility, as used intuitively in establishing trust. There are however many situations where individual reasoners can benefit from localized cooperation and derive a high utility from such yet retain a profound distrust. Often this is derived from past experience and some prior actions that have in the past significantly effected utility. Trust and its dual distrust are properties that persist and often decay slowly but yet can suddenly radically shift based upon apparently insignificant events somewhat akinned to catastrophe theory. We conjecture that utility monitoring, is the key to enabling trustworthiness. We will investigate architectural solutions that exploit our notions of agency, utility, and trustworthiness. Initial work in this regard has incorporated utility based reasoning into the BDI agent paradigm [11].

Novelty

Prior research typically adopts a given sensor management paradigm (query based, event-based or mobile agent-based). This RP in contrast advocates opportunistic and judicious choice of paradigm.

Previous research has optimisied streamed data by reordering of relational operators, filters and joins achieving significant optimisations in network placement [12]. This work assumed centralised processing. In contrast we consider aggregation over distributed, heteregeneous and resource-constrained devices. DFuse [13] and SBON [14] while addressing this assume powerful computational nodes. Our approach will incorporate dynamic role assignment [15] as a means of application distribution together with using collaborating agents in supporting migration, duplication and load balancing. [16] considers resource constrained device aggregation. To achieve truly power sensitive sensor networks acquisition issues of when, where, what and how data is sampled must be considered together with the manner (e.g. relational operator sequencing) by which such aggregation takes place.

Code and content dissemination strategies have tended to use flooding and gossiping techniques that blindly propagate content to all nodes failing to support selectivity of node updates determined by, for example update need, power availability or indeed node role. Ripple [17] utilised a publish and subscribe scheme while Tickle [18] periodically broadcasts software version metadate through polite gossiping, Deluge [19] extends Tickle to efficiently handle large data via a segmentation approach. Such approaches fail however to deliver content to targets based on need and furthermore fail to intelligently route to destinations. Target routing demands encoding the destination within the datum. Dissemination based upon perceived need is a key originality here. Reducing the size and number of communication datums and associated hops is crucial within such WSN constrained power envelopes.

The potential of agent based approaches has been recognised [8]. Systems have embraced mobile code/processes including Sensorware [20], Mate [21] and Pushpin [22]. Such systems adopt a weak agency injecting mobile reactive agents/code fragments into the network with viral network propagation. This research adopts a Belief Desire Intention (BDI) agent model whereby agents have a lightweight mental model of their environment which drives opportunistic reasoning at the individual and collective level and which is dynamically updated through perceptions and communication with other agents/nodes. It would seem therefore that due to the punitive cost of communication within WSNs that this model was ill-suited. However, due to the lightweight nature of the agent mental models communication can be driven by perceived utility (see WP 3.5) whereby the anticipated cost of communication would be offset by the utility payback deriving from more informed behaviour. [23][24] have explored mobile intentional agents in realising Autonomic Wireless Sensor Networks (AWSN). The availability of more powerful sensing platforms (Stargate, Sun Spot) further enables lightweight agents to be accommodated as evidenced by Agent Factory Micro Edition (AFME) [25][26]. AFME offers the smallest intentional agent software footprint available globally at a mere 84K smaller than other offerings like Cougaar MicroEdition (CougaarME) [27], 3APL-M [28] LEAP (Light Extensible Agent Platform [29] and MicroFIPA-OS [30].

Reference:

[1] D. C. et al., “Towards a sensor network architecture: lowering the waistline,” in HotOS-X: Proceedings of the 10th Hot Topics in Operating Systems Symposium, 2005, pp. 139–144.
[2] J. P. et al., “A unifying link abstraction for wireless sensor networks,” in SenSys’05: Proceedings of the 3rd ACM Conference on Embedded Networked Sensor Systems, 2005, pp. 67–89.
[3] S. Chandrasekaran, O. Cooper, A. Deshpande, M. J. Franklin, J. M. Hellerstein, W. Hong, S. Krishnamurthy, S. R. Madden, F. Reiss, and M. A. Shah, “Telegraphcq: continuous dataflow processing,” in SIGMOD ’03: Proceedings of the 2003 ACM SIGMOD international conference on Management of data. New York, NY, USA: ACM Press, 2003, pp. 668–668.
[4] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, and F. Silva, “Directed diffusion for wireless sensor networking,” IEEE/ACM Trans. Netw., vol. 11, no. 1, pp. 2–16, 2003.
[5] A. Goel and D. Estrin, “Simultaneous optimization for concave costs: single sink aggregation or single source buy-at-bulk,” in SODA ’03: Proceedings of the fourteenth annual ACM-SIAM symposium on Discrete algorithms. Philadelphia, PA, USA: Society for Industrial and Applied Mathematics, 2003, pp. 499–505.
[6] P. von Rickenbach and R. Wattenhofer, “Gathering correlated data in sensor networks,” in DIALM-POMC ’04: Proceedings of the 2004 joint workshop on Foundations of mobile computing. New York, NY, USA: ACM Press, 2004, pp. 60–66.
[7] L. Szumel and J. LeBrun, “Towards a mobile agent framework for sensor networks,” in Proceedings of the Emnets-II, 2005.
[8] A. Boulis, “Programming sensor networks with mobile agents,” in MDM ’05: Proceedings of the 6th international conference on Mobile data management. New York, NY, USA: ACM Press, 2005, pp. 252–256.
[9] T. D. Huynh, N. R. Jennings, and N. R. Shadbolt, “Certified reputation: how an agent can trust a stranger,” in AAMAS ’06: Proceedings of the fifth international joint conference on Autonomous agents and multiagent systems. New York, NY, USA: ACM Press, 2006, pp. 1217–1224.
[10] R. A. Shaikh, H. Jameel, S. Lee, S. Rajput, and Y. J. Song, “Trust management problem in distributed wireless sensor networks,” in RTCSA ’06: Proceedings of the 12th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications. Washington, DC, USA: IEEE Computer Society, 2006, pp. 411–414.
[11] S. Shen and G. OHare, “Wireless sensor networks: an energy-aware and utility-based bdi agent approach,” International Journal of Sensor Networks, vol. 2, 2006.
[12] C.-C. Han, R. Kumar, R. Shea, and M. Srivastava, “Sensor network software update management: a survey,” Int. J. Netw. Manag., vol. 15, no. 4, pp. 283–294, 2005.
[13] U. Ramachandran, R. Kumar, M. Wolenetz, B. Cooper, B. Agarwalla, J. Shin, P. Hutto, and A. Paul, “Dynamic data fusion for future sensor networks,” ACM Trans. Sen. Netw., vol. 2, no. 3, pp. 404–443, 2006.
[14] P. Pietzuch, J. Ledlie, J. Shneidman, M. Roussopoulos, M. Welsh, and M. Seltzer, “Networkaware operator placement for stream-processing systems,” in Proceedings of the 22nd International Conference on Data Engineering (ICDE’06). ACM Press, 2006.
[15] C. Frank and K. Römer, “Algorithms for generic role assignment in wireless sensor networks,” in SenSys ’05: Proceedings of the 3rd international conference on Embedded networked sensor systems. New York, NY, USA: ACM Press, 2005, pp. 230–242.
[16] S. Madden, M. J. Franklin, J. M. Hellerstein, and W. Hong, “The design of an acquisitional query processor for sensor networks,” in SIGMOD ’03: Proceedings of the 2003 ACM SIGMOD international conference on Management of data. New York, NY, USA: ACM Press, 2003, pp. 491–502.
[17] T. Stathopoulos, J. Heidemann, and D. Estrin, “A remote code update mechanism for wireless sensor networks,” Technical Report, vol. CENS-TR-30, 2003.
[18] P. Levis, S. Madden, J. Polastre, R. Szewczyk, K. Whitehouse, A. Woo, D. Gay, J. Hill, M. Welsh, E. Brewer, and D. Culler, “Tinyos: An operating system for wireless sensor networks,” in Ambient Intelligence. Springer-Verlag, 2004.
[19] J. Hui and D. Culler, “The dynamic behaviour of a data dissemination protocol for network programming at scale,” in Proceedings of 2nd International Conference on Embedded Networked Sensor Systems. ACM Press, 2004, pp. 81–94.
[20] A. Boulis, C.-C. Han, M. B., and Srivastava, “Design and implementation of a framework for efficient and progammable sensors networks,” in Proceedings of the First International Conference on Mobile Systems, Applications and Services. ACM Press, 2003, pp. 197–200.
[21] P. Levis and D. Culler, “Mate: a tiny virtual machine for sensor networks,” in ASPLOS-X: Proceedings of the 10th international conference on Architectural support for programming languages and operating systems. New York, NY, USA: ACM Press, 2002, pp. 85–95.
[22] J. Lifton, M. Broxton, and J. A. Paradiso, “Experiences and directions in pushpin computing,” in IPSN ’05: Proceedings of the 4th international symposium on Information processing in sensor networks. Piscataway, NJ, USA: IEEE Press, 2005, p. 57.
[23] D. Marsh, R. Tynan, D. O’Kane, and G. O’Hare, “Autonomic wireless sensor networks,” Engineering Applications of Artificial Intelligence Journal, 2004.
[24] R. Tynan, D. Marsh, D. O’Kane, and G. O’Hare, “Intelligent agents for wireless sensor networks,” in AAMAS 2005 - Proceedings of The Fourth International Joint Conference on Autonomous Agents and Multi Agent Systems. IEEE Computer Society, 2005.
[25] C. Muldoon, G. O’Hare, and M. O’Grady, “Collaborative agent tuning,” in ESAW 2005 - Proceedings of the Sixth International Workshop on Engineering Societies in the Agents’ World, 2005.
[26] C. Muldoon, G. OHare, and J. Bradley, “Towards reflective mobile agents for resourceconstrained mobile devices,” in AAMAS 2007: Proceedings of Sixth International Joint Conference on Autonomous Agents and Multi-Agent Systems. New York, NY: ACM Press, 2007.
[27] W. Wright and D. Moore, “Design considerations for multiagent systems on very small platforms,” in AAMAS 2003 - Proceedings of the 2nd International Joint Conference on Autonomous Agents and Multiagent Systems, 2003, pp. 1160–1161.
[28] F. Koch, “3apl-m platform for deliberative agents in mobile devices,” in AAMAS’05 - Proceedings of the Fourth international Joint Conference on Autonomous Agents and Multiagent Systems, 2005, pp. 153–154.
[29] M. Berger, S. Rusitschka, D. Toropov, M.Watzke, and M. Schichte, “Porting distributed agentmiddleware to small mobile devices,” in AAMAS 2002 - Proceedings of the Workshop on Ubiquitous Agents on Embedded, Wearable, and Mobile Devices held in conjunction with the joint conference on Autonomous Agents and Multi-Agent Systems, 2002.
[30] S. Poslad, H. Laamanen, R. Malaka, A. Nick, and A. Zipf, “Crumpet: Creation of user-friendly mobile services personalised for tourism,” in Proceeding of the Second IEE International Conference on 3G Mobile Communication Technologies, 2001.