Designing a Smart Farming Platform for Sustainable Decision Making

Abstract

Smart farming platforms (SFP’s) for pig livestock farming are of increasingly relevance to increase sustainable decision making and  enhancement of animal welfare. SFPs involve the whole supply chain and  integrate various types of data measured, thus enable data-driven solutions  using artificial intelligence. While there exists research about SFPs, issues  concerning data governance of SFPs are still lacking. Against this backdrop, we  develop a SFP for sustainable decision making with respect to data privacy and  data security. Our SFP integrates 4 sensor data sources (e.g., temperature  control system, and feeding stations), considers farmer characteristics (e.g.,  projects with pigs), and provides data-driven solutions (e.g., prediction of  animal welfare indicators). We report on the current process situation in pig livestock farming as well as on our concept of SFPs for sustainable decision  making. We also report on the evaluation of our SFP by validation against  defined requirements during the deployment phase.

Introduction

Sustainable digitalization is rapidly gaining importance in pig farming due to  requirements for efficient usage of resources and animal welfare monitoring. Animal  welfare monitoring is part of the strategy of the German government to improve  animal welfare without losing competitiveness [1]. However, animal welfare  monitoring is still in its infancies and current solutions are based on manual data  collections, which are sometimes evaluated only twice a year and increase production  costs due to increased manual labor [2, 3].

Research on animal welfare often use sensor data (e.g., video camera or

microphones) paired with data processing for monitoring and forecast solutions [4–6].  Because animal welfare is often animal-specific, these approaches capture and  process animal-specific data [7]. Therefore, Manteuffel and Schön propose a stress  level monitoring system using animal sounds [8]. In contrast, Matthews et al.  demonstrate a system to detect behavioral changes in pigs using depth video cameras  [9]. Also, equipment in bays e.g., feeding or drinking stations produces data to  monitor animal behavior [6]. Necessary sensors are already integrated into the  housing environment, e.g., feeding and drinking systems or scales. However,  especially in pig farming, these individual systems are not interlinked and data is  neither stored nor analyzed. This is often caused by economical limitations or technical aspects (e.g., appropriate interfaces to access data). Thus, there is a large  potential and responsibility for business informatics to aid farmers with a practical  data platform.

While animal welfare and economic aspects are often considered to be  incompatible, both issues can work together positively, e.g., prevention of diseases or  deaths [10]. Merging already existing data of farms, including sensor data, with data  of the whole supply chain in data platforms enable overall data analysis and data  processing [11].  

Smart farming platforms (SFP’s) can be considered as data platform with specific  focus on sensor based data collection and machine learning based supportive  functions.  As a result, these SFPs can provide functions to support decision making or to control machines [12–14].   While there already exists research about SFP’s for livestock farming, specific  solutions concerning data security and data privacy still lacking. Previous research  focused on independent supportive solutions. To monitor outdoor activities and to detect behavioral anomalies in cows, Taneja et al. propose an SFP for cow livestock  farming using interconnected pedometers. [15]. Also, Zhang et al. consider cows and  use attached sensors to cows and environment information to monitor activity of livestock’s (e.g., grazing) [16]. Ryu et al. demonstrate a concept for a SFP using  connected soil, air, and light sensors to enable mobile control of farm devices (e.g.,  fans, or sprinklers) [17]. Banhazi and Tscharke present a system to automatically  measure the live weight of pigs using image analysis without further data use [18]. In  contrast, Banhazi demonstrates a system with standalone sensors to monitor air  quality in stables automatically [19]. Also, Berckmans proposes a system to detect anomalies in the environment of broiler houses (e.g., feeding system failures or light  problems) [13]. However, these systems do not constitute a SFP, but can be a sub-part  of an SFP in terms of integrated data and provided functions. Rodriguez et al. propose  an open-source cloud data platform for crops, enable decision support for weed  control and traceability of growing processes [20]. Huang et al. show an SFP to  analyze supply and demand to support online sales and delivery of livestock products into supermarkets [21]. Additionally, they considered data security, such that users  combine a USB-key with standard login to access their data and supportive functions  [21].  

However, in our understanding, SFP’s integrate process related data as well as  sensory data to create an inter-company data platform and corresponding data-driven  solutions to enhance decision making. We address the following research question:  How to design an SFP concerning data governance for sustainable decision making?  Sustainable decision making refers to the decision making process, including definition of objectives and criteria’s, identifying consequences, and evaluation of  alternatives [22]. A decision is sustainable if the decision targets economic as well as  ecological objectives [23]. This paper has been divided into five parts. The first part  define requirements for an SFP using interviews with experts. The second section  describes our design of an SFP, split in data integration and data-driven solutions.  Subsequently, we report on our evaluation of the artifact using a real world  deployment and expert assessment. The final sections gives a brief discussion and  conclusion.