Deployment and Management of Cooperating Objects

Partners Involved

The DMCO cluster involves 13 members (5 associated): SICS (cluster leader), AICIA, ETHZ, NUIG, TUB, TUD, UDE, UCL, FBK (associated member), UCC (associated member), UBONN (associated member), UNITN (associated member), and UZL (associated member). 

Problem and Objectives

Based on the current state of the art, it is possible to observe a clear discrepancy between the expected and actual performance of systems of cooperating objects. This includes issues such as poor data yield, short system lifetime, and partial or complete failure. We identified two primary reasons at the origin of these issues:

• systems of cooperating objects are deeply embedded within the environment: the dynamics in real-world settings are generally unpredictable and difficult to simulate or replicate;
• devices employed in systems of cooperating objects are severely constrained: it is thus difficult to acquire the necessary visibility into the system state and operation.

Based on this analysis, the objective of the DMCO cluster is the development of a comprehensive framework (both at the conceptual level and with respect to concrete tools) to simplify the deployment and management of systems of cooperating objects.

Approach

To tackle the issues above, we aim at leveraging collaborations with domain experts and real-world experience, as intuitively shown in the following figure.

In collaboration with domain-experts, we gather real-world requirements in scenarios where systems of cooperating objects are to be deployed. Based on these, we develop a range of concepts and tools to simplify deployment and management. These cover all aspects involved, from the design to the implementation, optimization, and validation of the system.  We test our solutions in real-world scenarios, and ask the domain experts to provide feedback on the extent to which the system meets their requirement. We finally use this information to feed subsequent development iterations.

Activity Overview and Highlights

The research activities within the cluster are articulated within a framework spanning different stages of deployment and management. The framework is the result of a collaboration between DMCO and the FP7 project PLANET through DMCO member UDE, and is illustrated in the following picture. Specific information on the various research activities can be found at the end of the page, in the publication section.

In the initial planning stage, we provide solutions targeting the activities prior to the actual deployment. This includes the gathering and analysis of the requirements at stake, the initial coarse-grained system design, the study of the characteristics of the deployment environment, and initial performance estimations. For instance, in this context we are active in modeling the behavior of modern batteries, as cooperating objects are intended to operate autonomously, aiming at taking advantage of discharge patterns with smart communication functionality.  

Based on the initial design, users implement the necessary functionality. To this end, we are working in the design and implementation of dedicated programming paradigms, with the objective is to ease the development activity for programmers of cooperating objects. We target diverse scenarios and functionality, ranging from meta-programming techniques and abstractions for storage-centric settings, to the design and implementation of virtual machines for resource constrained devices.

In view of the deployment stage, users need to carefully plan the network layout to meet a range of requirements, e.g., in terms of communication reliability. In this context, the cluster is active in the area of deployment strategies. For instance, we are designing algorithms to decide on the placement of nodes to meet high-level connectivity requirements, such as a minimum number of redundant paths between given nodes.

Validation and optimization of the system design is fundamental before reaching the deployment stage. In this area, we are particularly investigating to what extent common assumptions in the state of the art match the characteristics of real-world scenarios. One example is our work on non-revisiting random walks techniques, where the objective is to understand the effect of realistic link dynamics on random walk querying techniques.

We test the solutions devised within the cluster in a number of on-site deployments. These represent a unique asset, in that they also allow us to gather precious feedback from domain experts. The scenarios at stake are very diverse, enabling a thorough assessment of our solutions, and include structural monitoring in heritage buildings, water monitoring in remote areas, bird tracking, road tunnel management and control, ambient assisted living, and chemical sensing in catastrophe situations.  

Following the first deployment, there is the need to diagnose and heal possible issues arising once the system is put in operation. We are thus working in the area of performance evaluation and debugging, as well as on self-healing mechanisms. Examples are work on visibility levels to allow programmers to trade resources for state information, and on network repair schemes using unmanned aerial vehicles and terrestrial robots.

Impact

The cluster has produced a number of scientific papers and prototype implementations. The complete list of research contributions can be found on the CONET publication page. Some examples, referring to some of the research activities exemplified above, are:

• K. Roemer, J. Ma. "PDA: Passive Distributed Assertions for Sensor Networks". In Int. Conf. on Information Processing in Sensor Networks (IPSN), 2009.

• N. Tsiftes, A. Dunkels, Z. He, T. Voigt. "Enabling Large-Scale Storage in Sensor Networks with the Coffee File System". In Int. Conf. on Information Processing in Sensor Networks (IPSN), 2009.

• N. Brouwers, K. Langendoen, and P. Corke. “Darjeeling: A Feature-rich VM for the Resource Poor.” In Int. Conf. on Embedded Networked Sensor Systems (SENSYS), 2009.

• L. Mottola. “Programming Storage-centric Sensor Networks with Squirrel”. In Int. Conf. on Information Processing in Sensor Networks (IPSN), 2010.

• M. Ceriotti, L. Mottola, G. Pietro Picco, A. L. Murphy, S. Guna, M. Corrà, M. Pozzi, D. Zonta, and P. Zanon. "Monitoring Heritage Buildings with Wireless Sensor Networks: The Torre Aquila Deployment". In Int. Conf. on Information Processing in Sensor Networks (IPSN), 2009. Best Paper Award.

• M Zuniga, C. Avin, and M. Hauswirth. “Querying Dynamic Wireless Sensor Networks with Non-Revisiting Random Walks”. In European Conference on Wireless Sensor Networks (EWSN), 2010.

• A. de Jong, M. Woehrle, and K. Langendoen, “MoMi – Model-based Diagnosis Middleware for Sensor Networks”. In Workshop on Middleware Tools, Services, and Run-time Support for Sensor Networks (MidSens), 2009.

• C. A. Boano, T. Voigt, N. Tsiftes, L. Mottola, K. Roemer, and M. A. Zuniga. "Making Sensornet MAC Protocols Robust Against Interference,'' European Conf. on Wireless Sensor Networks (EWSN), 2010.

• C. –Y. Shih and P. J. Marron. “COLA: Complexity-reduced Trilateration Approach for 3D Localization in Wireless Sensor Networks”. In Int. Conf. Sensor Technologies and Applications (SENSORCOMM), 2010.

• A. Bernauer, K. Roemer, and S. Santini.  "Threads2Events: An Automatic Code Generation Approach,'' ACM Workshop on Hot Topics in Embedded Networked Sensors (HotEmNets), 2010. 

• C.-K. Chau, F. Qin, S. Samir, M.H. Wahab, and Y. Yang. ``Harnessing Battery Recovery Effect in Sensor Networks,'' IEEE Journal on Selected Areas in Communications (JSAC), Special Issue on Simple Wireless Sensor Networking Solutions, 2009.

• D. Zonta, H. Wu, M. Pozzi, P.  Zanon, M. Ceriotti, L. Mottola, G. P. Picco, A. L. Murphy, S. Guna, and M. Corrà. "Wireless  Sensor Networks for Permanent Health Monitoring of Historic Buildings,'' SPIE International Journal on Smart Structures and Systems. Special Issue on Wireless Sensor Advances and Applications for Civil Infrastructure Monitoring. Volume 6, Issue 5-6, June 2010.

 For more information on the DMCO cluster, please refer to Luca Mottola.