Robotics for Precision Agriculture: from Kinematical Modeling and Control to Human-Multi-Robot Coordination

Modern agriculture faces increasing challenges driven by global population growth, labor shortages, and the need for more sustainable practices. To address these issues, robotics offers a compelling solution, particularly through the use of human-multi-robot teams. These systems can leverage the complementary strengths of humans and robots to perform complex tasks such as precision weeding, planting, and harvesting. However, realizing the full potential of such systems requires robust solutions for task allocation, navigation, and coordination. This work introduces two primary contributions to support the development of intelligent, collaborative, and autonomous systems for precision agriculture. The first contribution presents a Mixed-Integer Linear Programming (MILP) framework designed to optimize human-robot teamwork in agricultural contexts. Specifically, this framework is applied to a grape harvesting scenario, where the goal is to coordinate multiple service robots and a human worker efficiently. The proposed method minimizes task makespan, robot energy consumption, and idle time while incorporating essential constraints for human safety and comfort. A key innovation is the integration of dynamic velocity adjustments for robots during human-robot interactions to ensure smoother and safer collaboration. Moreover, a user-friendly interactive interface is developed to enable real-time adaptation of the task schedule and allocation based on human input and changing field conditions. This adaptability is crucial for real-world agricultural settings, where environmental variability and human feedback are significant factors. The MILP-based framework was validated through both simulation and real-world experiments. Simulations were conducted in a virtual vineyard environment to evaluate the system under realistic agricultural constraints. These were followed by laboratory tests using TurtleBots to demonstrate the system’s effectiveness in a controlled setting. The results showed that the framework could improve task efficiency, responsiveness, and human-robot coordination, thereby validating its potential for field deployment. The second contribution focuses on the development and control of a novel lightweight robot designed for autonomous weed and grass management in vineyards. This robot, developed during a Ph.D. internship at Vitirover, features a differential rear steering mechanism that enables effective navigation over rugged and uneven terrain. A kinematic model is derived to allow control using linear and angular velocity commands. To guide the robot to desired poses, a pose regulation controller is proposed. In addition, a dynamic optimal planner is introduced to manage precise navigation and weed targeting. The planner uses a multi-objective cost function with adaptive weights that shift priority based on whether weeds are detected within the robot’s field of view. This planning approach integrates kinematic constraints, environmental factors, and task-specific goals to ensure robust and efficient operation. This second system was validated through simulations in Gazebo and field experiments using custom-built hardware. The results confirmed the robot’s ability to autonomously manage weeds and navigate effectively in agricultural environments. Together, these contributions offer a unified framework for advancing robotics in precision agriculture. By addressing both autonomous field operations and human-robot collaboration, this work takes a step toward sustainable, scalable, and intelligent agricultural practices that respond to the complex demands of modern farming.