PROJECT SP3: Detection and actuation robotics systems for weed management (DARWEEM)
This project will address the issue of continuous monitoring using new early-perception algorithms that effectively detect and classify each weed seedling and are linked to treatment systems through the ISOBUS in robotics platforms. The project integrates machine learning combined with optimization techniques to build structures (deep neural networks, CNN, etc.) that best link the perception layer with the application layer. The following developments are complementary to the ones just described: a) new weed detection and classification tools; b) standardized communication protocols between the various systems comprising the robot; c) design, development and verification of autonomous navigation algorithms; and d) planning and supervision systems to coordinate robots while making sure that inspection and treatment are performed in a safe and effective manner. Considering the above, the following objectives have been set:
- Design, develop and evaluate automatic reconstruction and analysis systems in vineyards.
- Design, development and evaluation of automatic weed detection and recognition systems in maize and tomato fields. Connection with precision site-specific treatment.
- ISOBUS: Integration of all the systems on-board the platform through the standardized ISOBUS communication protocol
- Design, development and verification of autonomous and collaborative mobile inspection/treatment platforms (multi-robot teams).