This presentation was presented at the IEEE 5G Worldforum in a session 'Dialogues between 5G/B5G and Vertical Domains: AI for Intelligent Services. Several use cases in Food Systems that use 5G are presented of which the 'weed detection robot' in more detail. Enabling factors and recommendations for the use of 5G to create intelligent services using AI are discussed.
3. To explore
the potential
of nature to
improve the
quality of life
Wageningen mission
Digital Innovation
We bring the right partners together
who step-by-step develop digital
solutions, supported by state-of-
the-art knowledge on:
- data science
- information modelling and
management
- business modelling, governance,
and ethics
4. A vision on the application of AI in Food Systems
Successful Application of
Artificial Intelligence for
Sustainable Food Systems
Technically
robust
Socio-Economically
feasible
Socially acceptable
Ethically desirable
Human-centric
Fit-for-purpose
Explainable
Screening/auditing
Risk analysis
Power distribution
Labour effects
Human-robot
Inclusiveness
Discrimination
Tracking
5. Food Systems are increasingly being digitized
Smart Sensing
& monitoring
Smart Control
Smart Analysis
& Planning
6. Potential 5G connectivity benefits
Smart Sensing
& monitoring
Smart Control
Smart Analysis
& Planning
Edge computing for latency-sensitive
applications:
• storage and governance closer to the
farmer
• origin/certification information unlocked
to consumers.
• real-time product location
• monitoring systems: mobile apps, virtual and
augmented reality
• farmer's alerts with edge computing
• early warning in case of food incidents
• rescheduling i.c. unexpected food quality deviations
• product quality simulation based on ambient
conditions
• remote controllers for irrigation systems,
fertiliser systems, climate controllers,
harvesting systems, etc.
• achieve 99.999% accuracy with 5G
infrastructure with low risk
7. Exploring 5G use cases for Food Systems
7
Aggregation level
Farm system
Food system
Cross-sectoralrural area
Cyber-physical
Management cycle
Monitoring
Analysis & planning
Control
Decision-making level
Individual
(business/consumer)
Supply chain (food
integrity)
Public (society/
government)
Use case typology
8. 5G smart farming applications
8
This Photo by Unknown Author is licensed under CC BY-NC
Weed Detection
Robot
Real-time haulm
killing drone
Water quality
measurements 5Groningen
5G Rural Integrated
Testbed (5GRIT)
5G Range for
remote areas
Smart Farm
Innovation Valleys
9. Weed detection robot (Netherlands)
9
Autonomous weed robot: cameras and spot sprayer
attached to the vehicle to control plant specific
weeds.
Detection is a power- and data consuming operation
that requires costly hardware.
Offloading computational operations and data
storage to a nearby infrastructure (cloud computing)
reduces costs and energy consumption to extend
operational time in the field. Photo 1 Weed detection robot (“Agrointelli Robotti
spuit plant-specifiek | LandbouwMechanisatie,”
n.d.)
Performance
• > 90% of weeds accurately detected and sprayed while < 5% of the main crop
sugar beet was sprayed incorrectly.
• Uplink 120 Mbps and 25 ms latency (upload images, processing images and
downloading locations of plants)
10. Weed detection robot – use case analysis
10
Aggregation level Farm system
Food system
Cross-sectoralrural area
Cyber-Physical
Management cycle
Monitoring
Analysis & planning
Control
Decision-making
level
Individual
(business/consumer)
Supply chain (food
integrity)
Public (society/
government)
11. 11
33 use case projects
>> 12 future 5G use cases
Source: www.iof2020.eu/trials
Internet of
Food and
Farm 2020
Innovation Action:
2017 - 2020
30 M€ funding by
DG-CNCT/AGRI
Large-scale
uptake of IoT in
the European
farming and food
sector
12. Increased use of open (public) data, private farmer data and data from
service providers (e.g. contractors, soil laboratories, agribusiness)
Ensuring interoperability by using standards & solid cyber security measures
New business models for mobile operators (network sharing and slicing) and
the monetisation of data for all parties involved
Sponsoring mobile traffic for rural areas to ensure fairness and inclusiveness
Policy changes aiming at increasing the quality of life in rural areas and
continued Public Private Partnership (PPP) initiatives
Ecosystem building as key enabler with local Digital Innovation Hubs (DIHs)
as a one-stop shop for farmers, investors, incubators, technology providers
Enabling factors for 5G uptake of in agri-food
12
13. Recommendations to leapfrog development
13
Scenario
I 5G implementations
available
II IoT implementations
available, without
5G
III No Smart Farming
and Food production
applied
Aggregation level Decision-making level
Management cycle
Extend from farm
level to food system
or cross-sectoral/rural
Close management
cycle by including
control mechanisms
Extend individual business
to supply chain and/or
public decision-making
Start at the farm
level to explore the
benefits of 5G
Start at monitoring
level, gradually
expand to analysis
and control
Start at individual business
decision-making level to
explore the benefits of 5G
Identify actors at
different levels that
want to start with
smart farming and
food production
Start at the
monitoring level
with most promising
technology
Start at the individual
business decision-making
level with most promising
technology
14. Thanks for your
attention!
14
Dr. J. (Sjaak) Wolfert
Strategic Sr. Scientist Digital Innovation in Agri-Food
Wageningen University & Research
sjaak.wolfert@wur.nl
Editor's Notes
1
In our vision successful application of AI for Sustainable Food Systems can only be achieved if the solutions are
technically robust - building human-centric AI software and hardware that is fit-for-purpose, explainabe, while implementing adequate screening and auditing of algorithms
socio-economically feasible for those implementing and using it, including appropriate risk analysis, evaluating possible shifts in power distributions and effects on labour and
if it is developed and used in an ethically desirable and socially acceptable way including human-robot relations, inclusive, exclusive or discriminatory algorithms, tracking technologies and so forth.
These three pillars are often mutually intertwined, and a trade-off between the pillars can expected.
A close collaboration of researchers and practitioners of multiple disciplines is therefore required.
The domain of agriculture, food and the environment is increasingly being digitized through the introduction of all kind of smart devices and software. We would expect that this is something for technical specialists and engineers involving tech-oriented scientists. However, I want to argue that this is much more a social experiment requiring social scientists.
Referring to the figure above I distinguish 4 application areas in which digital data increasingly play a role:
Digital data is becoming more important for (1) decision-making for businesses at any level of the agri-food supply chain; from farmers, through logistic providers to consumers.
The same data is essential for (2) food integrity, providing assurance to consumers and other stakeholders about the safety, authenticity and quality of food.
(3) Public decision-making for societal challenges such as food security, climate change, healthy food and nutrition could also tap into these data instead of using separate censuses and statistics which are usually lagging behind.
Finally, this digitization is driven by fast developments in (4) science and technology (S&T), such as Artificial Intelligence, Internet of Things, Blockchain, etc. At the same time, advancements in data science also heavily rely on the data that is being generated by the application of data-driven research; simply put: there is no data science without data.
In conclusion, the same digital objects can be used for multiple application areas and through that these areas become increasingly intertwined.
While application areas 1 (business decision-making) and 2 (food integrity) are mainly driven by business, big steps in development are already being made. However, for application areas 3 (public decision-making) and 4 (S&T) this is more difficult. Access to private data from businesses and consumers is challenging for various reasons such as lack of incentives and business models or the risk of data misuse. However, in the end private businesses are also heavily relying on public decision-making and developments in S&T.
Therefore, transdisciplinary research is needed and becoming also more feasible as a result of the digitization. This means that knowledge, concepts, methods from various disciplines and multiple data sources from real practice are being integrated in interaction with actual stakeholders in living labs.
This means that all relevant stakeholders have to collaborate with each other in in real-life contexts developing new digital solutions in iterative cycles based on real(-time) data. Digitization is therefore much more a social experiment requiring substantial involvement of social scientists.
If you are further interested in this, I would like to refer to the project IoF2020 and SmartAgriHubs in which we try to put this approach into practice.
Weed Detection Robot (arable)
Real-time haulm killing from drone to task map to spray application (arable & livestock case)
Smart Farm Innovation Valleys (horticulture)
5G Range Agribusiness and Smart Farming for Remote Areas (arable & livestock)
5G Rural Integrated Testbed project (5GRIT) (arable & livestock)
Water quality measurements 5Groningen (Future for 5G) (arable & livestock)
An autonomous potato weed control initiative leveraging 5G uses cameras on a vehicle and a deep learning algorithm on an on-board computer, recognising the weeds.
A spot sprayer attached to the vehicle controls plant specific weeds. The detection is a power and data consuming operation that requires costly hardware.
Offloading these operations (computational/data storage) to a nearby infrastructure (cloud computing) reduces costs and energy consumption of electric drive robots to extend operational time in the field.
Current 3G and 4G communications technologies lack sufficient bandwidth and need lower latency to perform real-time operations in the field (taking pictures with cameras, sending the pictures to a cloud server, performing deep learning calculations, sending position of recognised weeds back to vehicle, controlling weed). A field demonstration with pre-5G connectivity was successful. More than 90% of the volunteer potato was accurately detected and sprayed while less than 5% of the main crop sugar beet was sprayed incorrectly. The outcomes realised were an uplink bandwidth of 120 Mbps and a latency of 250 ms (upload images, processing images and downloading locations of plants) (Wageningen University & Research, 2019).
CPM cycle demonstrated by Weed Detection robot case
Monitoring: camera’s for detection, Sensoring & analysis: Weed detection robot with deep learning algorithm on an on-board computer, recognising the weeds, Control: actionable spray to control weeds (instead of manual farmer actions).
In total we work on 33 use case projects all over Europe. Next slides provide an illustrative example of one use case demonstrating the multi-actor approach.
The future of 5G agriculture will depend upon a number of enabling factors including interoperability, data governance and security, new business models, policy changes, and innovative ecosystems.
Increased use of open (public) data, private farmer data and data from service providers (e.g. contractors, soil laboratories, agribusiness).
Ensuring interoperability by using standards is key to the 5G developments and devices that exchange data on a smart farm need solid cyber security measures.
Uptake of 5G includes new business models for mobile operators (network sharing and slicing) and the monetisation of data for all parties involved, because data is a key asset.
Sponsoring mobile traffic for rural areas to ensure all have a fair share is an example of such a new business model.
Policy changes should be aiming at increasing the quality of life in rural areas and continued Public Private Partnership (PPP) initiatives.
Ecosystem building is a key enabler with local digital innovation hubs that act as a one-stop shop for farmers, investors, incubators, technology providers, etc. to get connected.
@sjaak, deze nog gebruiken?
Recommendations to leapfrog development for three scenarios along three dimensions