Sensor Cloud Integration: An Agent-Based Workbench for On-the-Fly Senor-Data Analysis

Synopsis:

This research project focuses on integrating wireless sensor networks into grid and cloud computing systems, and gives an agent-based workbench for facilitating on-the-fly sensor-data analysis on top of these computing systems.

Although the emergent popularity of cloud computing has allowed users in SME (small and medium-sized enterprise) and SOHO (small office/home office) to construct their business-specific computing systems with cloud services, the cloud-computing users still encounter some practical difficulties in particular when handling a large amount of sensor data: remote job execution, on-the-fly data analysis with parallelization, and sensor-data delivery.

This research targets frost protection as an example application where tree-fruit growers use a temperature sensor network so as to observe and to possibly predict every overnight transition of their orchard temperatures. While they can use various cloud services including remote storages, computing power, and even programming tools, they still need to address by themselves the following three problems: (1) automating the execution of a temperature-prediction program upon arrivals of new temperature data, (2) accelerating the program execution using cloud-offered programming tools, and (3) forwarding sensor data to the program and saving outputs into file servers. Without proper solutions, tree-fruit growers would end up with repeatedly checking the current orchard temperatures.


Figure 1. Orchard in Wapato, WA


Figure 2. Sensor-Data Analysis on Grid/Cloud


Figure 3. Temperature Interpolation and Prediction

The project addresses these problems by providing an agent-based workbench that glues individual cloud services together with a collection of three software tools, each named AgentTeamwork-Lite, MASS, and Connector:

  1. AgentTeamwork-Lite dispatches and moves a user job to the most-fitted computing resources, using mobile agents that spontaneously roll down on a steep slope of a computing-resource potential field, as if they were electrons sliding on an electrical potential field in physics. Such a computing-resource potential field is maintained among computing nodes by exchanging their resource information with neighboring nodes, which relieves users from even identifying appropriate servers.


    Figure 4. Field-Based Job Dispatch and Migration


    Figure 5. System Layers

  2. The MASS library enables a user to compose their application of distributed array elements and multi-agents, each coded with entity-based programming where all computation is enclosed in each element or agent, (i.e., an entity) as well as all communication is scheduled as periodical data deliveries from one to other entities using their relative indices. This model unleashes an application from accessing entities in for-loops, and thus eases dynamic allocation of entities to multiple computing nodes and multiple CPU cores.


    Figure 6. Parallel Execution with MASS Library


    Figure 7. Example Code using MASS Library

  3. The Connector toolkit converts sensor and multimedia data originated from various network protocols to stream-oriented files that are automatically pumped to and even redirected to a migrating user job. File-based and automatic data redirection removes protocol/source-specific data-retrieval code from a user program, and thus improves the code maintainability and reusability.


    Figure 8. Connector

The project is to complete these software tools, to demonstrate its applicability to on-the-fly sensor-data analysis, to make the system available to our collaborators (in frost protection, wireless network, and video analysis), and to disseminate the deliverables to the public.

Members:

Summer 2010 Lab Members Field Trip to Wapato, WA
Summer 2011 Lab Members UW Business Plan Competition 2011 Pacrim 2011 Conference, Victoria, BC

Documents:

Publications

  1. Qinghong Shang, Munehiro Fukuda, Michael B. Dillnecourt, Lubomir Bic, Automatic Resource Management in Multi-site Mobile Computing (Invited Paper) In Proc. of the 5th Int'l Conference on Mobile Computing and Ubiquitous Networking - ICMU 2010, Seattle, WA, pages 146-152, April 26-28, 2010
  2. Munehiro Fukukda, Shuichi Kurabayashi, Jeremy Hall, Yasushi Kiyoki, Morphing Parallelization Strategy to Support On-the-Fly Video Analysis In Proc. of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications - PDPTA 2010, Las Vegas, July 12 - 15, 2010
  3. Jose Melchor, Munehiro Fukuda, A Design of Flexible Data Channels for Sensor-Cloud Integration In Proc. of 21st International Conference on System Engineering - ICSEng 2011, pages 251-256, Las Vegas, NV, August 16-18, 2011
  4. Munehiro Fukuda, Agent-Based Workbench for On-the-Fly Sensor-Data Analysis In Proc. of 2011 IEEE Pacific Rim Conference on Communications, Computers and Signal Prcessing - PACRIM'11, pages 333-339, Victoria, BC, Canada, August 24-26, 2011
  5. John Emau, Timothy Chuang, Munehiro Fukuda, A Multi-Process Library for Multi-Agent and Spatial Simulation. In Proc. of 2011 IEEE Pacific Rim Conference on Communications, Computers and Signal Prcessing - PACRIM'11, pages 369-376, Victoria, BC, Canada, August 24-26, 2011
  6. Elad Mazurek, Munehiro Fukuda, A Parallelization of Orchard Temperature Predicting Programs In Proc. of 2011 IEEE Pacific Rim Conference on Communications, Computers and Signal Prcessing - PACRIM'11, pages 179-184, Victoria, BC, Canada, August 24-26, 2011
  7. Tomoaki Matsuno, Satoshi Kushioka, Jungo Imahara, Munehiro Fukuda, Tadanori Mizuno, Hiroshi Mineno, "Development of Agri-Environmental Visulalization and Control System using Spatial Interpolation", In Proc. of DICOMO 2012, pages 2129-2136, Ishikawa, Japan, July 4 - 6, received the excellent paper award.
  8. Masashi Sakurada, Munehiro Fukuda, An RSSI-based Error Correction Applied to Estimated Sensor Locations, In Proc. of 2013 IEEE Pacific Rim Conference on Communications, Computers and Signal Prcessing - PACRIM'13, pages 58-63, Victoria, BC, Canada, August 27-29, 2013
  9. Somu Jabayalan, Munehiro Fukuda, Field-Based Job Dispatch and Migration, In Proc. of 2013 IEEE Pacific Rim Conference on Communications, Computers and Signal Prcessing - PACRIM'13, pages 52-57, Victoria, BC, Canada, August 27-29, 2013
  10. Yuta Aoki, Tadao Oishi, Masaki Bandai, Munehiro Fukuda, Takashi Watanabe, "Load Balancing of Multi-Sink Sensor Networks with Asymmetric Topology and Traffic Patterns," Transaction of The Institute of the Electronics, Information and Communication Engineers, accepted on June 24, 2013, to appear

Presentations

"Sensor Grid Integration: An Agent-Based Workbench for On-the-Fly Sensor-Data Analysis"
at Faculty Research Prsentation, Computing and Software Systems, University of Washington, Bothell 		Colloquium April 20, 2011

Manuals

  1. MASS: Parallel-Computing Library for Multi-Agent Spatial Simulation, May 7th, 2010