Digital Twin and Machine-Learning for Optimized Sensor-driven Precision Agriculture Utilizing Autonomous Ground and Aerial Delivery

Background

The overall guiding motivation is to provide useful tools to enable rapid path planning for autonomous vehicle operators in real-time and to train operators in large surface area food systems. This work develops a digital twin and machine-learning optimization framework for model problems combining:

Fluid-dynamics of released objects ranging from powders to encapsulated packets and sensors
Ground-based vehicle or aircraft (unmanned or manned) dynamics
Energy-efficient path planning of multiple ground or air vehicles
Lidar-based and hyperspectral data extraction for agricultural mapping for planning and assessment.

Goals

Specifically, we will develop clean and easy-to-use digital tools that are repeatedly in demand by food system researchers such as:

Data collection
Compression and assimilation from the environment
Rapid ODE and PDE solvers
Neural nets and genetic algorithms with embedded detailed models of fundamental physics (fluid flow, heat transfer, stress analysis, electromagnetism, optics, robotics, etc.)

Impact

The framework is designed to enable digital twin type technologies to run at much faster rates, in order to enable machine-learning to optimize the planning.