The world population is estimated to grow and reach about 10 billions in 2050, significantly raising the demand for food.
Especially in developing countries, where the economy is almost entirely based on agriculture, the increasing demand for food is exasperated by climate changes and by scarce infrastructures and farm tools.
New approaches and technologies are needed to mitigate the effects of climate changes and spreading plant diseases, to produce sustainable food, and eventually meet the UN zero hunger goal.
Currently adopted technologies for crop monitoring mostly rely on human operators, who plan the monitoring operations and analyze the collected data.
Unfortunately, these approaches fail in providing timely support and disease detection in large areas, allowing the pest/disease agent to expand, leading to epidemic outbreaks (e.g. rusts), before being detected.
Differently from current solutions, this project proposes an autonomous monitoring system to continuously monitor the crops, detect diseases, and, based on historical data, provide predictions about the most critical crops.
We consider the use of a squad of aerial drones, which monitor crops and autonomously adapt their trajectories, upon detection of ongoing diseases or detection of critical anomalies.
The drones collect data and directly analyze them using a special-purpose Artificial Intelligence (AI) algorithm.
In detail, we provide and test the AI models for plant disease detection, which use images of plant leaves; we design the algorithms and protocols for path planning and task assignment for monitoring and data collection; and finally we integrate a forecast module to predict critical zones and adapt the drone trajectories accordingly.
We envision that our system may help farmers to increase food production, by reducing disease spreading thanks to timely detection and intervention.
Our approach is innovative because it provides a completely automated system that guides drone trajectories, detects possible critical zones, on the basis of historical data, and dynamically adapts the drone trajectories to provide refined monitoring of critical zones and possible infected crops.
This is innovative with respect to the current state of the art in many aspects.
- Some previous proposals dealing with crop monitoring rely on reactive approaches, based on the only observation of anomalies in the field of interest. Instead, our approach aims at continuously monitoring the crops, allowing prediction of possible critical zones and detection of infected crops (i.e., both reactive and proactive approach).
-Also, current approaches mostly rely on human operators for data analysis and disease detection. Our approach aims at carrying Machine learning algorithms on drones to automatically detect possible diseases, reducing human intervention.
-Current approaches focus on drones for precision agriculture use only single, remotely controlled drones. We are advancing the state of the art by integrating a cooperative and autonomous squad of drones.
- Current approaches focused on smartphones and IoT sensors are hard to implement in developing countries, due to the scarce availability of resources and infrastructure for communication. Instead, a team of drones can communicate in a multi-hop manner (i.e., they do not require ground infrastructure) and can provide timely intervention in big areas also with few drones, thanks to their high mobility (i.e., they have lower deployment cost).
We believe that such an approach will provide broader impacts and can inspire future monitoring systems for food security. Agencies and governments may provide these systems to developing countries, increasing food production, thanks to the timely detection of diseases, and prompt intervention.
Also, the produced algorithms for task assignment, path planning, and autonomous coordination of drones, will advance the state-of-art algorithms for squads of drones. These algorithms in fact may be used in any similar scenarios: patrolling, monitoring, target inspection, and so on.