Creating High-performance/Low-cost Ambient Sensor Cloud System Using OpenFS (Open Field Server) for High-throughput Phenotyping Masayuki Hirafuji1, 2 Hideo Yoichi1, Takuji Kiura1, Keiko Matsumoto1, Tokihiro Fukatsu1, Kei Tanaka1, Yukinori Shibuya1, Atsushi Itoh1, Hirohisa Nesumi1, Norihiro Hoshi1, Seishi Ninomiya3, J. Adinarayana4, D. Sudharsan4, Yasunori Saito5, Kazuki Kobayashi5, Takanobu Suzuki6 1
National Agriculture and Food Research Organization, Japan (E-mail:
[email protected]) 2 University of Tsukuba, Tsukuba, Japan 3 University of Tokyo, Tokyo, Japan 4 Indian Institute of Technology Bombay, India 5 Shinshu University, Nagano, Japan 6 Nagano Prefecture Agricultural Experiment Station, Nagano, Japan Abstract: Recently high-performance sequencers can read whole DNA sequence in short-time. On the other hand, phenotypic information is too short to analyze genomic data. Such phenotypic data will be collected in various environmental conditions (i.e. many places) by long-term observations for individual plants. It can be technologically realized by using low-cost multifunctional sensor networks and free cloud services instantly. We developed OpenFS (Open Field Server) employing open-source hardware. Also free cloud data service such as Twitter is employed to share observed data. Concept of ambient sensor cloud system is proposed by using both OpenFS and cloud computing. Keywords: Sensor Network, Field Server, Cloud, Environment, Monitoring, Phenotyping
1. INTRODUCTION Recently high-throughput sequencing technology is enabling to collect whole data of plant genome in short time [1]. However, data of phenotype of varieties is still short comparing to the genotypic data because there have not been automatic tools to measure both environments and phenotypic data simultaneously for long term. For example, huge phenotypic data manually was collected to dissect variation for flowering time with a set of 5000 recombinant inbred lines, where a million plants were assayed in eight environments, since flowering time is a complex trait that controls adaptation of plants to their local environment [2]. If we can use genotypic data, phenotypic data and environmental data for each individual, relationships between gene and traits can be automatically identified. Thus we need to collect enormous data in fields [3]. Although we have developed multi-functional device for sensor network, Field Server which can monitor many kinds of information simultaneously [4, 5], it is still too high-cost and maintenance is too hard for such purposes although the cost is 1000-10,000USD. We need to deploy sensor nodes in the world widely to collect massive data for high-performance phenotyping, and then key points are more rugged, longer life span, lower cost and easier to monitor distribution of soil moisture at many points and physiological status of individual plant such as stem moisture and LAI, much easier devices are needed to deploy them vastly. Also we should propagate the sensor network technologies widely in agriculture and developing countries to collect data vastly: it must be extremely low-cost and simple. Also, we have other problems: 1. Functions and sensors should be changed easily for
difference of environments, plants and areas. Firmware of sensor nodes should be customized by ordinal programmers or users easily without expensive software tool kits. We developed OpenFS (Open Field Server) employing open-source hardware, Arduino, to solve these problems. Also free cloud data service such as Twitter is employed to share observed data easily. Concept of ambient sensor cloud system is proposed by combining both concept of OpenFS and cloud computing. 2.
2. OPEN-FS OpenFS (Open Field Server) is a kind of Field Server without a camera case and Field Server Engine. For OpenFS, we employed open source hardware, Arduino, instead of Field Server Engine. Advantages of the open source hardware are low-cost, rich software libraries and much information about bugs and applications. Although function of Arduino is much fewer than Field Server Engine (FSE), Arduino with add-on boards (shields) such as Ethernet shield and X-Bee (ZigBee) shield is enough to use as simple sensor nodes. Moreover, we can design our own original shields to make it compatible to conventional FSE. Thus, combining open source hard ware, open source software and resource of various add-on boards, we can develop high-performance sensor-network and propagate sensor network technology in agriculture widely. Field Server employed Wi-Fi for a strategic reason to propagate Wi-Fi technology [6], so OpenFS also keeps this strategy. Additionally we developed a mother board for Arduino board and Ethernet Shield (Fig. 1). Arduino board on the mother board can controls power of an
Ethernet Shield and an external Wi-Fi router to save power consumption. The mother board can restart the on-board devices: these kinds of devices may be freeze for long-term-use. Mean power consumption can be below 100mW by using the mother board and a power control program. As the result, daily power consumption could be lower than 2.4Wh, which could be generated by small solar panels as shown in Fig. 2. As the result OpenFS will work for low-cost extension sensor nodes in conventional Field Server sensor networks.
3. AMBIENT SENSOR CLOUD Ubiquitous sensor network enabled sensor data collection at many points in real-time, and collected data was stored on data storage such as PCs, database servers or Web servers. However it is hard to maintain all of the systems forever. As for Field Servers, we have been operating Agent systems to collect data, Web servers to share the collected data in real-time, and applications [5]. Such services are operated by researchers for their research projects in fact. It is very hard to maintain the services for long-term sustainably after the research projects finished. Free cloud services such as Twitter and Pchube can play a role of data storage and some applications. Already there are libraries and techniques to use API of Twitter and Pachube on software
development tool (Arduino IDE) and Webs. We employed Twitter for data storage for OpenFS (Fig. 3). Besides, data archive cloud service for Twitter, Twilog, is also used simultaneously. This is an example to increase robustness using different cloud services for a redundant storage system.
4. DISCUSSHION AND CONCLUSION Proposed ambient sensor cloud system composed of OpenFS, Twitter and Twilog was tested in India as a first trial. Then we deploy at fields of NARO such as an orange orchard in Japan. Deployments at the sites were quite easy comparing to conventional deployment which was very hard for sever working environment such as high-temperature, unstable electricity and insufficient engineering time. We will be able to collect enormous data at many sites, if time of deployment and maintenance at other sites can be shorten like these trails. Power consumption of Wi-Fi devices was small sufficiently comparing to ZigBee devices. Recently cost of most Wi-Fi routers is 20-60USD, which is lower than ZigBee devices. We could use DD-WRT [7] to utilize more useful functions for Wi-Fi routers in free. Moreover, we could eliminate cost of media conversion between Wi-Fi (and Ethernet) and ZegBee by using only Wi-Fi routers.
Fig. 1 circuit of the mother board of OpenFS
As the result, cost of an OpenFS can be lower than 400USD including sensors such as soil/plant moisture, temperature and LAI and an irrigation controller without mass-production. Based on our recent experiences such as the large earthquake in East Japan, mobile cell phones cannot be available. However optical fiber for the Internet and cloud services such as Twitter could serve normally at many sites and many people could get important information. Truly the Internet and cloud services were robust against the disaster. Combination of the Internet, Wi-Fi and cloud technology is the most reliable and the best solution from a view point of tolerance. Combination of OpenFS, open source technologies, the Internet, Wi-Fi and cloud computing will be able to realize high-throughput phenotyping.
ACKNOWLEDGEMENTS This research has been carried out partially in the SCOPE (Stragetic Information and Communication Promotion Programme) project (#102304002) supported by MIC (Ministry of Internal Affairs and Communications) in High Definition Image Live Agri-information Japan. Trials in India was carried out as Strategic Japanese-Indian Cooperative Programme on “Multidisciplinary Research Field, which combines Information and Communications Technology with Other Fields”, Geo-ICT and Sensor Network based Decision Support Systems in Agriculture and Environment Assessment. LED Lighting with IR sensor
Wi-Fi
Solar Panel 700mm
100mm
Soil Moisture Sensor Fig.2 Field Twitter composed by OpenFS
Battery Voltage
Inside Temperature
Soil Moiture (Raw)
Fig.3 An examples of sensor data tweeted by OpneFS on Twitter.
REFERENCES [1] Shendure, J., H. Ji, “Next-generation DNA sequencing”, Nature Biotechnology, 26, pp. 1135-1145, 2008. [2] Buckler, E. S. et al., “The Genetic Architecture of Maize Flowering Time”, Vol. 325, Science, pp. 714-718, 2009. [3]Lee, W. S., V. Alchanatis, C. Yang, M. Hirafuji, D. Moshou, C. Li, “Sensing technologies for precision specialty crop production”, Computers and Electronics in Agriculture, Vol.4, 1, pp. 2-33, 2010. [4] Hirafuji, M. and T. Fukatsu, “Architecture of Field Monitoring Servers”, Proc. of the Third Asian Conference for Information Technology in Agriculture, pp. 405-409, 2002. [5] Fukastu, T., M. Hirafuji, T. Kiura, “Web-based Sensor Network with Flexible Management by Agent System”, Advances in Practical Multi-Agent Systems, SCI325, Springrt-Verlag, Berlin Heidelberg, pp. 415-424, 2010. [6] Hirafuji, M., “Creating Comfortable, Amazing, Exciting and Diverse Lives with CYFARS (CYber FARmerS) and Agricultural Virtual Corporation”, Proc. of the Second Asian Conference for Information Technology in Agriculture, pp. 424-431, 2000. [7] http://www.dd-wrt.com/site/index
