igreen: co-creative mobile services in agriculture - a knowledge management perspective
TRANSCRIPT
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 1
Andreas Dengel
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 2
Agenda
Some words about the world‘s food situations
The iGreen project
Agricultural knowledge management – two perspectives
Consolidating knowledge and decision support services
Summary and conclusions
Crop forecasting
Logistic planning
Soil quality map adjustment
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 3
Agriculture must provide food for an extra 2 billion people but 80% of extra agricultural food can be provided only through yield increase!
Source: (1) UN, World population to exceed 9 billion by 2050. http://www.un.org/esa/population/publications/wpp2008/pressrelease.pdf, 2008 (2) N. Alexandratos, J. Bruinsma, G. Boedeker, J. Schmidhuber, S. Broca, P. Shettym, and MG Ottaviani. World agriculture: towards 2030/2050. Interim Report. Prospects For Food, Nutrition, Agriculture and Major Commodity Groups, 2006.
9 billion people by 2050!
Extra agricultural food production
By increasing yield By increasing arable land
100%
≤20% ≥80%
Too many factors High complexity!
?
2010 2050
+43%
World demand for cereal (bn tonnes2)
2,1
3,0
-‐0.2%
2050 2010
Amount of arable land in developed countries
(mn hectares)
625 575
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 4
IT can help to improve and economize methods, procedures and agricultural technologies to produce more efficiently and save resources
¨ Analysis of high-‐resolution images from space: Controlled specialized satellites take pictures of agricultural region all over the world
¨ Field robots will take over some agricultural tasks: Equipped vehicles (not only for harvesting) are staffed with sensors to measure the state of crop and soil
¨ Precision farming: Linked and controlled by GPS, seed, fertilizer and pesticides are yielded exactly where they are needed
¨ Digital soil quality maps: Continuous measure lead to updated guidelines helping to plan the supply of nutrients or predict harvest
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 5
What are the
ideas of the iGreen project?
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 6
Consul'ng services Interests of the social community Interests of the agricultural sector
Das Bildelement mit der Beziehungs-ID rId7 wurde in der Datei nicht gefunden. Farmers / agricultural service
supply agency
iGreen builds on an alliance of 23 partners from science, business and public institutions
private public
Applica4on Management
EPP
e.v .
(na4onwide)
iis
(interna'onal) (na'onal) (regional)
Connec4vity
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 7
iGreen strikes a new path in public-‐private knowledge management having a focus in agriculture
¨ The goal of iGreen is to develop location-‐bases services and knowledge sharing networks for combining distributed, heterogeneous public as well as private information sources
¨ Built on that, we aim at the development of mobile decision assistants using Web services for a decentralized support of
energy-‐efficient, economic, environmental-‐adapted and collaboratively-‐organized
production and planning processes
¨ There are many high-‐potential application fields for the iGreen platform, such as agriculture, forestry, water supply and distribution, urban development and landscaping, or nature conservation
¨ in iGreen we exemplary focus on crop production
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 8
In crop production there is whole bunch of relevant questions having a space-‐time relationship that have to be answered for making decisions
When should we apply what kind of fertilizer and to which dose?
Which kind of plant species should we cultivate at what location?
How much pesticides should we apply at
what time?
When is the best time for the harvest?
What market prices may we expect?
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 9
Public country-‐wide institutions in Germany act as competence centers and provide best-‐practice advice about crop production for farmers
Information about how to combat pest and plant diseases
(best time, conditions and strategies)
Competence Center
Compendium about Best-‐Practice Crop Protection
Regional Data About Fields and Lots
Data Base about Crop Experiment Results
News about Market Development
Support for Data Management and Model Generation
Advise in Grade and Seed Selection
Consulting about Fertilizer Combination and Use
Tools for Prognoses and Statistical Data Analysis
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 10
Some examples about the services from the competence centers
Competence Center
Exposition Regional Factor
Gradient
++ =Risc Map
2009 2011
Pest Immigration
Weather Station Information (Point⇒Zone)
⇒ ⇒
Data Interpolation Methods (MR, Kriging, spline, ....)
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 11
?
Yield depends on many factors – Experts and Decision Support Systems help farmers to optimize yields
Ecological factors
• Soil quality • Weather condition (rainfall)
• Sun exposition
Social factors
• Transfer of skills and roles
• Security of land tenure
• Access to land
Economical factors
• Fuel, fertilizers, pesticides prices
• Product selling prices (Market)
• …
Political factors
• Agricultural subventions
• Quotas
• Other restrictions
Yield
+ Ecological factors + Economical factors + Political factors + Social factors + Model(s) ________________ Optimized Yield
So how to get all of these factors combined?
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 12
iGreen profits from recent technological advances that, in the intended combination, will strongly influence agricultural progresses
¨ Public information as well as product and expert knowledge are increasingly digitized and available via the internet, such as different kinds of maps or weather data
Soil Quality Estimation
Methods
Thematical Maps
¨ Accessing the internet is possible from almost all locations via mobile devices
¨ GPS devices for positioning are easy to buy or are even part of a mobile phone
¨ Agricultural engineering provides more and more software interfaces for the automatic situation-‐adaptive control (on-‐board terminals, sensor technology, …
So we are going to establish a Service and Knowledge Network
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 13
Competence Center
Public
Private
The iGreen Planning System combines knowledge subjects and information objects for a participative public-‐private knowledge management
Data Provision Geo Catalogue Agro Databases
Feed
back
Feedback Prognoses
Global Optimization
Local Optimization Free Service for Farmers
Planning System
Farmers Location-‐based Decision Assistance
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 14
Flexible Data Management is a challenge
¨ The only condition is individual data ownership
-‐ local filing combined with controlled interchange within the iGreen network
Before you start talking about sensor data, please provide me first an option to file, manage
and employ the data
Farmer and Contractor Marx
“ “
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 15
Public-‐private
knowledge management
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 16
Knowledge Management is a process for improving organizational capabilities by better use of individual and collective knowledge resources
1. Knowledge Goals: Define all capabilities an organization should build on
2. Knowledge Identification: Identify internal and external knowledge of the organization
3. Knowledge Acquisition: Critical capabilities must be bought or otherwise obtained
4. Knowledge Development: Produce new internal and external knowledge (individual & collective level)
5. Knowledge Dissemination: Define who should know what and at what level of detail, and how the organization can support this distribution process
6. Knowledge Utilization: Productive deployment of organizational knowledge in the business/production process of the organization
7. Knowledge Preservation: Identify valuable knowledge, store it, and regularly integrate it into the organizational knowledge base
8. Knowledge Controlling: Compare initial knowledge goals with results of organizational knowledge magt.
Source: G.J.B. Probst, S. Raub, and K. Romhardt. Wissen managen. Gabler, 1997. ISBN 3409193170.
Definition
2
Identify knowledge
4
Develop knowledge
5 Distribute knowledge
6 Use knowledge
7 Preserve knowledge
8 Control knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Feedback Knowledge!Management!
Cycle!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 17
In terms of knowledge management, both experts and farmers have their own knowledge management cycle
Increase Yields
Develop DSS
2
Identify knowledge
4
Develop knowledge
5 Distribute knowledge
6 Use knowledge
7 Preserve knowledge
8 Control knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Feedback
Decision Support System (DSS)
Define knowledge
Goals
Disseminate knowledge 5
3 Acquire knowledge
4
2
6
1
8
7
Use knowledge
Preserve knowledge
Control knowledge
Identify knowledge
Develop knowledge
Feedback
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 18
In terms of knowledge management, both experts and farmers have their own knowledge management cycle
Define knowledge
Goals
Disseminate knowledge 5
3 Acquire knowledge
4
2
6
1
8
7
Use knowledge
Preserve knowledge
Identify knowledge
Develop knowledge
2
Identify knowledge
4
Develop knowledge
5 Distribute knowledge
6 Use knowledge
7 Preserve knowledge
8 Control knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Feedback
Decision Support System (DSS)
Increase Yields
Develop DSS
Feedback
Control knowledge
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 19
In terms of knowledge management, both experts and farmers have their own knowledge management cycle
Define knowledge
Goals
Disseminate knowledge 5
3 Acquire knowledge
4
1 Develop
knowledge
4
Develop knowledge
6 Use knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Decision Support System (DSS)
Increase Yields
Develop DSS
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 20
Increase Yields
In terms of knowledge management, both experts and farmers have their own knowledge management cycle
Define knowledge
Goals
Disseminate knowledge 5
3 Acquire knowledge
4
1 Develop
knowledge
4
Develop knowledge
6 Use knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Soil Type
SSIISSLsLLLTTMo
Soil quality map from the 1950’s (Low definition and outdated)
No better decision support !
No better map material!
Not Possible!
No better advice
Decision Support System (DSS)
Develop DSS
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 21
Source: R. Grisso, M. Alley, and P. McCellan. Precision farming tools: yield monitor. Precision Farming, pages 442–502, 2003.
GPS Technology combined with agricultural sensors to measure the yield
Available agricultural technology allows farmers to derive accurate and up to date soil quality maps
Mass Flow Sensor Moisture Sensor
GPS Receiver
Task Computer User Interface
Soil quality map from the 1950’s : • Outdated • Low resolution
Soil quality map from precision ag.: • Up to date • High resolution
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 22
Increase Yields
A collaborative knowledge management approach can lead to better agricultural decision support
Define knowledge
Goals
Disseminate knowledge 5
3 Acquire knowledge
4
1 Develop
knowledge
4
Develop knowledge
6 Use knowledge
3
Acquire knowledge
1
Define knowledge
Goals
Better Decision Support
Better DSS
Possible
Yield in t/ha
0 -‐22,01 – 44.01 – 66.01 – 8>8
Yield maps (High definition & up-‐to-‐date) Better
map material!
Win-‐Win Situation
Develop DSS
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 23
Farmers can collaboratively contribute to the acquisition of better geo-‐data and get in return better decision support
Partial collaboration: Data is only shared with the experts
Full collaboration: Data is shared within the community
Geo-‐Data
Information Supply
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 24
Scenario planning allows
crop forecast
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 25
The collaborative contribution was implemented in applications for biomass and logistics planning
Biomass Planning
Goal • Computing the optimal biomass yield based on: – list of fields – production plan – soil quality – weather conditions
Resources • Experts provide models and weather data • Farmers provide accurate soil quality maps using precision agriculture and a production plan
Approach • Soils quality maps provided by a farmer are used to improve results but remain confidential
• Models are improved and benefit the whole community
Examples of Applications and Evaluations
Partial collaboration!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 26
Tasks and their dependencies of Biomass-‐Yield-‐Models (BYM) are transformed into a scientific workflow
n Advantages of scientific workflows: • Data flow approach based on visual programming • Modularity and reusability • Provenance information to better interpret results and
debug errors
Computational steps for scientific simulations or data-‐analysis steps
Source: B. Ludäscher, I. Altintas, C. Berkley, D. Higgins, E. Jaeger, M. Jones, E.A. Lee, J. Tao, and Y. Zhao. Scientific workflow management and the Kepler system: Research articles. Concurr. Comput. : Pract. Exper., 18:1039-‐1065, August 2006. ISSN 1532-‐0626. doi: 10.1002/cpe.v18:10.
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 27
Based on various public sources we have developed a Crop Forecast System aiming at supporting the farmer in making decisions
* Zentralstelle der Länder für EDV-‐gestützte Entscheidungshilfen und Programme im Pflanzenschutz
Rain
Soil Quality Data
Soil Quality
Intermediate Result
Crop Forecast Result Map
Result Table
Result Chart
Agricltural Crop Land Online Rheinland-‐Pfalz (FLOrlp)
Region
Biomass Yield Models (BYM)
BYM
Filter
Pre-‐given Goal
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 28
A Mashup of maps and tabular information provides a intuitive platform for biomass planning
Results for fields
Results for subfields
Ecological or conventional yield
3 rainfall scenarios (dry, normal, wet)
Excel and Google Earth export
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 29
In Google Earth subfield-‐related information may be accessed via any mobile device
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 30
© [email protected] - 2008
Exporting the information to an Excel spreadsheet (with Macros) supports realizing a production plan
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 31
We have tested the forecast on a farm by evaluating four tours*
512 550
667
459
1.215
443
651
-‐2%
-‐19%
-‐15%
-‐10%
Tour 4 Tour 3 Tour 2 Tour 1
Biomass (t)
Real Forecast
Weed infestations
One field not be harvested
* 16 fields to be harvested; avg. dist. betw. field and silo: 9.5 km; average exploitable acreage: 2.44 ha
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 32
Logistic planning allows to remarkably save fuel
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 33
The collaborative contribution was implemented in applications for biomass and logistics planning
Biomass Planning
Goal • Computing the optimal biomass yield based on: – list of fields – production plan – soil quality – weather conditions
Resources • Experts provide models and weather data • Farmers provide accurate soil quality maps using precision agriculture and a production plan
Approach • Soils quality maps provided by a farmer are used to improve results but remain confidential
• Models are improved and benefit the whole community
Examples of Applications and Evaluations
Logistics Planning
• Computing the optimal logistics plan, i.e., route and costs for: – harvester(s) – tractor(s) with trailers
• Experts provide models • Farmers provide accurate GPS-‐Tracks and meta-‐information about their routes using precision agriculture and number of harvesters and tractors available
• GPS-‐Tracks and meta-‐information provided by a farmer are sanitized and shared with the whole community
• Models are improved and benefit the whole community
Full collaboration!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 34
Current available route guidance systems do not provide information about road narrowing nor bridge weights limits
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 35
This way, we also developed a first collaborative open-‐source routing system for utilities vehicles (CRUV)
Vehicle type (car, lorry, tractor)
-‐ Road types: Motorway, highway, country road, field path with priorities (preferred, normal, avoid, forbidden)
-‐ Tunnels and bridges with priorities
Vehicle length, weight, height, width
Rules based on user meta-‐information stored in Open Street Maps
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 36
CRUV: 186 km Map24: 186 km
Compared to Map24, there is no difference in quality in traditional route planning
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 37
GoogleMaps: 1,5 km CRUV: 2,8 km
However, CRUV reveals its advantages when including the additional information, e.g. with bridges and maximal weight allowed
Google Maps does not support the kind of query!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 38
A logistic planer computes transportation costs based on the results from the CRUV and biomass planner
Route between POI based on driving distance Slope (of the road)
Point of interest (ex: Biogas plant)
List of Fields (< d): -‐ Size -‐ Driving distance -‐ Crop -‐ Yield -‐ Transportation costs
Field repartition with respect to the driving distance
Max. driving distance d between field and POI
Overview with Biomass and costs
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 39
We have tested the forecast on a farm by evaluating four tours*
512 550
667
459
1.215
443
651
-‐2%
-‐19%
-‐15%
-‐10%
Tour 4 Tour 3 Tour 2 Tour 1
Biomass (t)
Real Forecast
Weed infestations
One field not be harvested
* 16 fields to be harvested; avg. dist. betw. field and silo: 9.5 km; average exploitable acreage: 2.44 ha
6.714
14.077
5.086
6.320 6.542
8.003
6.816
Tour 2 Tour 1
+8% +57%
+4%
-‐3%
Tour 4 Tour 3
Real Forecast
Transportation costs (EUR)
Higher fuel consumption due to dry soil
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 40
The goals of iGreen are very relevant for many social systems
¨ iGreen follows an approach leading to more accurate and geographically differentiated agricultural forecast by
-‐ leveraging existing farmers GPS-‐based precision agriculture technology
-‐ collaboratively acquiring accurate, high-‐resolution, and up-‐to-‐dare geo-‐data
-‐ allowing public agricultural institutions, in return, to use this geo-‐data to provide farmers with better decision support tools
¨ Biomass Planner
-‐ A decision support system relying on this approach to provide better biomass planning
-‐ Both the biomass model and soil quality maps are improved but no data is shared with the community. Only the improved model benefits the community
¨ Logistic Planner -‐ A decision support system relying on this approach to
provide better logistics planning
-‐ Both the model and the geo-‐data are shared with the community
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 41
Moreover,
existing soil quality maps may be continuously improved
So, we are going to close th
e!
Knowledge Management!
Cycle!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 42
Remember that forecast is based on the data coming from existing soil quality maps!
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 43
40
33 44
53
61
Digital soil quality maps provided by state government allow farmers to get an overview about the quality of their fields
Additional information may be accessed via the lot No.
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 44
Agricultural engineering provides relevant location-‐based data
Recording of location, crop, humidity, and fuel consumption:
Humidity Sensor
Melt Flow Sensor
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 45
CROP (t/ha) to from
Location-‐based recording of crop data leads to an up-‐date of soil quality maps
Recorded on 29-‐09-‐2008 Average Value 12,23 t/ha Try Solids 100,40 t Humidity 73,55 %
HISTOGRAM
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 46
Decision support in crop production provides important contributions for increasing efficiency, saving resources, and reducing environmental impact
¨ Winter wheat is the most important crop in Germany having a total acreage of 3 Mio hectares. Experts estimate that each hectare of winter wheat requires about 200 kg of nitrate per year. Nitrate production causes about 50% of energy consumption in crop production
¨ The German agriculture may …
... lead to possible energy saving potential of 30.000 tons of nitrate per year
... avoid climate-‐relevant emissions of 257,000 tons of CO2 annually
... reduces the amount of nitrate-‐containing nutriments in bodies of water
¨ Using location-‐adaptive fertilizer dispersion for wheat production based on improved soil quality maps , it was possible to save ca. 5% of nitrates without reducing the amount of crop
Exemplary result (10 farmers in the north of the German state Rhineland-‐Palatinate):
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 47
The goals of iGreen are very relevant for many social systems
¨ Food production is not a task just of agriculture, but a central aim of the global society
¨ In case of crisis, food security has to be guaranteed by both, federal and private organizations (this is part of the food precaution law and food ensuring law in Germany)
Long-‐term Public-‐Private Partnerships are of major importance
¨ iGreen use cases and demonstrators prove the new options based on real application scenarios
¨ iGreen SDK provides general and fundamental components as open source
¨ iGreen documents (processes, scenarios, interfaces) provide excellent guidelines, e.g. who communicates with whom ,why, and with what tools
© German Research Center for Artificial Intelligence - [email protected] 2012 - Page 48
Thank you, questions are welcome!