The placement of grazing animals in vineyards requires additional support to the animal husbandry activities. Such support must include the monitoring and the conditioning of animal's location and behavior, specially their feeding posture. With such a system, it is possible to allow sheep to graze in cultivated areas (e.g. vineyards, orchards) without endangering them. This paper proposes an animal behavior monitoring platform, based on IoT technologies. It includes an IoT local network to gather data from animals and a cloud platform, with processing and storage capabilities, to autonomously shepherd ovine within vineyard areas. The cloud platform also incorporates machine learning features, allowing the extraction of relevant information from the data gathered by the IoT network. Thus, besides the platform description, some results are presented regarding the machine learning platform. Namely, this platform was evaluated for detecting and defining conditions respecting animal's posture, with preliminary promising results. Since several algorithms were tested, this paper includes a comparison of those algorithms.

Intelligent farming is one of the vast range of applications covered by the Internet of Things concept. Notwithstanding, such applications present specific requirements and constraints that are dependent on their purpose. A practical case on which that is particularly relevant is the SheepIT project, where an automated IoT-based system controls grazing sheep within vineyards, guaranteeing that they do not threaten cultures. Due to its rigid requirements, particularly regarding the deployment of the Wireless Sensor Network, Machine-2-Machine communications and necessary interactions with a computational platform available through the Internet, Internet Protocol-based solutions are not suitable. Consequently, a customized communication stack has been developed, that intends to meet the project requirements, from the physical to the Application Layers. Although it has been developed considering the SheepIT requirements, its use may be extended to more generic intelligent farming applications, since most of the requirements are directly related with the farming environment. This paper reviews the proposed stack and details the recent developments. Particularly, we focused on Internet of Things/Machine-2-Machine interaction, describing the design and deployment of a gateway that addresses the SheepIT service requirements. Additionally, and complementary to previously published results, we evaluate the gateway performance and show its feasibility and scalability in a real scenario.

Weed control in vineyards demands regular interventions that currently consist of the use of machinery, such as plows and brush-cutters, and the application of herbicides. These methods have several drawbacks, including cost, chemical pollution, and the emission of greenhouse gases. The use of animals to weed vineyards, usually ovines, is an ancestral, environmentally friendly, and sustainable practice that was abandoned because of the scarcity and cost of shepherds, which were essential for preventing animals from damaging the vines and grapes. The SheepIT project was developed to automate the role of human shepherds, by monitoring and conditioning the behaviour of grazing animals. Additionally, the data collected in real-time can be used for improving the efficiency of the whole process, e.g., by detecting abnormal situations such as health conditions or attacks and manage the weeding areas. This paper presents a comprehensive set of field-test results, obtained with the SheepIT infrastructure, addressing several dimensions, from the animals’ well-being and their impact on the cultures, to technical aspects, such as system autonomy. The results show that the core objectives of the project have been attained and that it is feasible to use this system, at an industrial scale, in vineyards.

IoT covers several application domains, each of them with different features, requirements and constraints. Precision agriculture is one of such domains, having very specific and rigid requirements, including typically a WSN, M2M communications and interaction with computational platforms available through the Internet. Despite its popularity, IP-based communication stacks are not suitable for many of these kind of applications, mainly because of the overhead associated. In this paper we propose a new M2M Communication Stack, designed to fit the Sheep IT project requirements, but developed with more generic applications in the area of intelligent farming in mind. Along with the stack definition, this paper also includes experimental results showing the feasibility of the proposed stack.

The use of animals, specially ovines, for controlling infesting species in crops is an ancient method that has several advantages with respect to the use of chemical and/or mechanical techniques. Shepherds required to supervise animals, are getting increasingly less available and are costly. Over the last years, technologies for automatic animal supervision have been investigated, typically including devices to carry out localization, apply stimulus to the animals and, in some cases, report health data. Ovines have a relatively small size, which constrains the maximum weight that they can carry on, and consequently, imply that a practical system must be highly energy-efficient, to allow building light and small devices with enough autonomy. This paper presents a new system for sheep grazing control, developed in the scope of the SheepIT project. The focus of the paper is on the global communication architecture, which is a fundamental aspect for the overall system efficiency. The paper also includes simulation and experimental results that address the scalability and energy consumption aspects of the system.