Veröffentlichung von "Establishing resilient AI applications in agriculture by redundancy and graceful degradation: Two use cases" / Publication of "Establishing resilient AI applications in agriculture by redundancy and graceful degradation: Two use cases" [30.10.24]
Paper auf der AgEng 2024 Conference unter der Beteiligung von Sebastian Bökle und Fachgebietsleiter JProf. Anthony Stein erschienen. - Paper published at AgEng 2024 Conference with the participation of Sebastian Bökle and our head of department JProf. Anthony Stein.
ENGLISH VERSION BELOW.
Ein weiteres Paper ist aus der Arbeit des KI-Fachgebiets erschienen.
Hier kommen die Hard-Facts zur Veröffentlichung.
Titel:
Establishing resilient AI applications in agriculture by redundancy and graceful degradation: Two use cases
Autoren:
S. Bökle, M. Müller, W. Keil, H. W. Griepentrog, A. Stein
Venue:
AgEng 2024
Abstract:
Digital agriculture encompasses the development of AI-supported machines and processes. The resulting intelligent agricultural technologies have the potential to increase process and resource efficiency and to simplify farmers’ workflow of crosschecking the validity of machinery adjustments. In this paper, the benefits of a digitised and AI-assisted sowing process of cereals are presented. This case study focuses on (i) the prediction of the remaining working time and (ii) the prediction and order placement of in-time delivery of seed refill. As farming processes are subject to variable weather conditions, the time windows for their application are narrow, which holds for the seeding process in particular. On the other hand, the integration of AI for digital support of the seeding process may lead to possible constraints due to vulnerable networks and unreliable internet connectivity. This raises the necessity of a high level of technical resilience for the investigated use case of AI-supported seeding. To counteract this connectivity issue, redundant fall-back options for scenarios (i) and (ii) have been developed. More precisely, for (i), we follow a graceful degradation approach. While not maintaining the predictive accuracy level of the cloud-based prediction using the full amount of telemetry data, it still enables a seamless process continuation based on cached data of the most recent calculations. For scenario (ii) another redundant fall-back solution was developed, where the prediction is computed on a farm server proactively caching a permanent status report, traced from the former cloud-based predictions. Thus, the principle behind is not a parallel calculation but a permanently occurring poll, sent to the corresponding digital service, assuring permanent access to the required data for on-edge computations during fieldwork. Within the context of food production as critical infrastructure, we consider the thoughtful incorporation of technical resilience in AI-enhanced agriculture as paramount to ensure a fail-safe food production system.
Unter folgendem Link kommen Sie zur gesamten Übersicht der bisherig veröffentlichten Forschungsarbeiten des Fachgebiets: ki-agrartechnik.uni-hohenheim.de/veroeffentlichungen
--
Another paper has been published with the participation of the AI department.
Here are the hard facts about the publication.
Title:
Establishing resilient AI applications in agriculture by redundancy and graceful degradation: Two use cases
Authors:Autoren:
S. Bökle, M. Müller, W. Keil, H. W. Griepentrog, A. Stein
Venue:
AgEng 2024
Summary:Digital agriculture encompasses the development of AI-supported machines and processes. The resulting intelligent agricultural technologies have the potential to increase process and resource efficiency and to simplify farmers’ workflow of crosschecking the validity of machinery adjustments. In this paper, the benefits of a digitised and AI-assisted sowing process of cereals are presented. This case study focuses on (i) the prediction of the remaining working time and (ii) the prediction and order placement of in-time delivery of seed refill. As farming processes are subject to variable weather conditions, the time windows for their application are narrow, which holds for the seeding process in particular. On the other hand, the integration of AI for digital support of the seeding process may lead to possible constraints due to vulnerable networks and unreliable internet connectivity. This raises the necessity of a high level of technical resilience for the investigated use case of AI-supported seeding. To counteract this connectivity issue, redundant fall-back options for scenarios (i) and (ii) have been developed. More precisely, for (i), we follow a graceful degradation approach. While not maintaining the predictive accuracy level of the cloud-based prediction using the full amount of telemetry data, it still enables a seamless process continuation based on cached data of the most recent calculations. For scenario (ii) another redundant fall-back solution was developed, where the prediction is computed on a farm server proactively caching a permanent status report, traced from the former cloud-based predictions. Thus, the principle behind is not a parallel calculation but a permanently occurring poll, sent to the corresponding digital service, assuring permanent access to the required data for on-edge computations during fieldwork. Within the context of food production as critical infrastructure, we consider the thoughtful incorporation of technical resilience in AI-enhanced agriculture as paramount to ensure a fail-safe food production system.
The following link will take you to the complete overview of the research work published to date by the department: ki-agrartechnik.uni-hohenheim.de/veroeffentlichungen
