International Graduate School Cottbus / IHP microelectronics Im Technologiepark Frankfurt (Oder) Germany IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved Fault Tolerant Event Specification in Heterogeneous Sensor Networks Ortmann, Steffen

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 2/23 Outline Introduction & Motivation Related work Shortcomings and open issues Fault tolerant event specification Conclusion

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 3/23 Motivation

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 4/23 Introduction Ubiquitous systems are ambient intelligent environments build by cooperating autonomous devices Computing devices are to be embedded on everyday object Watching and serving us at any place and any time Supposed to substitute todays computers and information technology Reliable and fault tolerant ubiquitous systems are potentially capable of executing mission- and safety-critical applications Healthcare- and structural monitoring Homeland security Embedded systems Avionic and deep space applications

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 5/23 Sensor networks Sensor networks are one of the first real world examples of ubiquitous systems Tiny autonomous devices that are assembled to fulfill common tasks Structure: Main challenge: Devices and systems are prone to failures Low cost devices, rare resources, strict energy constraints

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 6/23 Tmote Sky sensor node from Moteiv Main Features: 250kbps 2.4GHz IEEE Wireless Transceiver 8MHz Texas Instruments MSP430 microcontroller 10k RAM, 48k Flash Integrated ADC, DAC, Supply Voltage Supervisor, DMA Controller Onboard antenna with 50m range indoors / 125m range outdoors Integrated Humidity, Temperature, and Light sensors Programming and data collection via USB Ultra low current consumption Fast wakeup from sleep (<6μs)

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 7/23 Related Work Most reliability enhancing approaches for sensor networks focus on data gathering and data transmission Events are predefined states based on certain measurements Usually defined by threshold values Exploit the effect of redundancy on mean time to failure Strongly depends on the density in the network Main approach: collective distributed data evaluation by voting Neighbored nodes compare their results to decide about events Many different voting algorithms are presented so far

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 8/23 Distributed event evaluation

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 9/23 Voting Based on redundant devices in certain areas of the network Cluster, position-based, n-hop neighborhood etc. Majority Voting [1] All nodes within the region of event possess the same weights Fusion center analyzes and combines all values Distance Weighted Voting [2] Voting weight decreases with distance to the center of the event Confidence Weighted Voting [2] Grants higher weights to sensors that are more likely to be correct Every nodes assigns a confidence value Best implemented in the TIBFIT [3] protocol

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 10/23 Shortcomings and open issues Voting is sufficient enough to provide an enhanced reliability Currently done on predefined event regions Not adaptable to different tasks and network conditions Consider heterogeneous sensing capabilities Almost all sensor network applications are handmade and customized Take care on energy dissemination Varying tasks demand different overhead for fault tolerance Exploit reactive algorithms that vote on demand only! Vision: adaptable multi-tasking sensor networks Miscellaneous fine-grained multi-event detection

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 11/23 Event Specification Language for Sensor Networks Idea: Enable complete events specifications by an uniform description language Get rid of custom-built sensor networks! Combine heterogeneous sensing capabilities Enable more precise and complex event detection capabilities Fine-grained configuration of fault tolerant event evaluation Configure voting conditions explicitly for any single event Specify execution intervals and associate appropriate event handlers Online configuration of sensor networks without physical access to every node!

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 12/23 element Structure of an events specification …

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 13/23 element List sensing capabilities e.g.,, etc. Configure corresponding threshold values Exact threshold values as element Scopes of threshold values by or Correlate threshold values by logic operations,,, etc. element is analyzed to a Boolean value during evaluation of sensor readings

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 14/23 example

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 15/23 element (1) Customizes preconditions for distributed event evaluation Precisely configures conditions for voting Determines which other devices are allowed to vote Defines the legal size of the event evaluation region All nodes within this area are allowed to vote element defines a radius around initiating sensor node Using quantifying elements like Other preconditions are to be considered too e.g. all nodes within 1-hop neighbourhood

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 16/23 element (2) Number of necessary voting devices can be fixed or limited Stated by element Enables n-modular redundancy Specification of further abort criteria (called Exceptions) Listed by the element Deadline criteria Keeps timing constraints for safety-critical applications! Other criteria imaginable All listed criteria can be concatenated by logic operations

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 17/23 example 5 3 3

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 18/23 element Configuration of demand-oriented execution intervals Precisely adaptation to varying requirements Implicitly considers energy consumption of the sensor node Manages active and sleep periods of the sensor node Can be quantified by acceptable time periods or exact time slots Example: 60

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 19/23 element Connects procedures to an event Procedures are called event handlers element holds a list of event handlers Every event handler links a certain procedure Attribute id holds the respective identifier All listed handlers are successively executed if an event occurs Example:

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 20/23 Conclusion Reliable and fault tolerant sensor networks are in great demand Enable mission- and safety-critical applications Current approaches and solutions revealed several shortcomings Idea: Define events by an uniform event specification language Regards heterogeneous sensing capabilities Allows for fine-grained event-related fault tolerance Provides miscellaneous task execution Improves maintenance capabilities and enables online configuration of sensor networks

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 21/23 Outlook Finish definition of event specification language Implement pre-parser for event specifications Parses specification into tree that can be send through the network Implement interpreter for the sensor node side Using network (OMNet++) and algorithm simulator (Castalia) Comprehensive test procedures on simulator Different network density Different event specification containing varying voting conditions Measure and compare energy consumption

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 22/23 References [1] B. Krishnamachari and S. S. Iyengar. Efcient and fault-tolerant feature extraction in sensor networks. In 2nd Workshop on Information Processing in Sensor Networks, IPSN03, Palo Alto, California, April 2003 [2]T. Sun, L.-J. Chen, C.-C. Han, and M. Gerla. Reliable sensor networks for planet exploration. In L.-J. Chen, editor, Proc. IEEE Networking, Sensing and Control, pages 816–821, Tucson, USA, 2005 [3]M. Krasniewski, P. Varadharajan, B. Rabeler, S. Bagchi, and Y. Hu. Tibt: trust index based fault tolerance for arbitrary data faults in sensor networks. In Proc. International Conference on Dependable Systems and Networks DSN 2005, pages 672–681, 2005

IHP Im Technologiepark Frankfurt (Oder) Germany © All rights reserved 23/23 Discussion Thanks for your attention. Any questions?