SARA-ESMERA: Autonomous robotic platform for the precision application of plant protection treatments in steep-slope vineyards

Edge AI
Edge AI
IA predictiva y simulación inteligente
IA predictiva y simulación inteligente
Sector
Agriculture, livestock farming and fisheries
Aplication area
Sustainability and territorial innovation
Geographical area
International

SARA was a precision agriculture solution developed by CTIC together with SVMAC under the European project ESMERA (European SMEs Robotics Application). The platform combined artificial intelligence and robotics to apply plant protection treatments precisely and locally in steep-slope vineyards, a type of crop with particular operational limitations.

The project tackled the challenge of applying plant protection treatments in steep-slope vineyards, which account for between 10% and 12% of Europe's wine-growing area and produce some of the most valuable wines. In these vineyards, with rows 90 to 150 cm wide, treatments were usually applied with small tractors fitted with air-blast systems, which caused high product losses and soil compaction.

The objectives of the project were:

  • To develop a robotic platform capable of precisely applying plant protection treatments in both conventional and steep-slope vineyards.
  • To incorporate an intelligent system for detecting vines and leaf density that would adjust the amount of treatment applied to each plant.
  • To ensure the safety of people, animals, vines, and the platform itself through obstacle detection.

The areas on which SARA was built were:

  • Precision robotic application: an articulated-arm actuator applied the treatment locally, adapting to the geometry of conventional and steep-slope vineyards.
  • Intelligent vine detection: a system analyzed the images captured by a camera on board the platform to identify the vine, its relative position, and its leaf density (no leaves, low, medium, or high density), sending this information to the actuator to adjust the applied dose. To train this computer vision system, a Sim2Real strategy was used, generating synthetic data in a virtual environment, which sped up the training of the algorithm.
  • Obstacle detection: a front-facing camera analyzed the surface ahead of the vehicle to identify obstacles within a safety perimeter.

SVMAC coordinated the project, with CTIC's collaboration in the development of SARA. The first verification tests of the system were carried out at CTIC RuralTech, a sandbox for rural innovation located in the Peón Valley (Villaviciosa, Asturias), with two buildings and 17,000 m² of land dedicated to experimentation in a controlled real-world environment.

Partners

More information

  • Status: Completed
  • Completion date: 01/05/2020 - 30/04/2021
  • Coordinador: SVMAC
  • Infraestructura singular: RuralTech

Funded by

European Union's Horizon 2020 Program, through the ESMERA (European SMEs Robotics Application) project