Multi-Agent Collaborative Framework for Automated Agriculture

The use of internet-connected devices, especially small multi-rotor Unmanned Aerial Vehicles (UAVs), in scientific data gathering and applications is quite widespread. But due to limited intervention capability, the UAVs alone fail to automate agricultural tasks completely. Thereby, we propose a centralized framework capable of handling a heterogeneous mixture of UAVs and UGVs to cater to the needs of automating agriculture efficiently. The framework’s core is a novel heuristic decision module that creates new tasks by visually analyzing the farm and solves a vehicle routing problem to allocate it to agents optimally. It is also equipped with supporting modules to monitor their operation and, in case of failures, help them recover autonomously based on the task and agent assessment. The framework is used in three significant agricultural applications, namely yield prediction and drought stress detection in a simulated environment using ROS and Gazebo, and 3D mapping of a real farm. These applications demonstrate the use of the multi-agent collaborative framework in identifying agricultural tasks on a farm and executing them.