Development of an Autonomous Laser-Based Robotic System for Sustainable Weed Management in Vidalia Onions

View the project report

Commodities

• Vegetables: onions

Practices

• Crop Production: Weed Management

In the United States, onion farming predominantly relies on the transplanting of onions, as seeded onions are highly vulnerable to the spread of weeds. Growers are compelled to opt for transplanting despite the high costs, due to the absence of organic and reliable weeding practices. Weeds compete for vital resources in the soil and attract insects and diseases, which can severely impact crop health and yield. Traditionally, weed management has relied on manual weeding and the application of chemical herbicides. Manual weeding is labor-intensive, expensive, and often harms the crops it is meant to protect. Conversely, while effective at controlling weeds, the use of herbicides introduces environmental risks such as soil and water contamination, damage to non-target species, and the development of herbicide-resistant weed populations. As demand grows for organic and sustainable farming, chemical weed control is becoming less viable. In response to these challenges, this research introduces an autonomous weeding robot system designed specifically for onions, utilizing laser technology to precisely target and eliminate weeds. This robotic system has three main objectives. First, to accurately detect weed plants in onion crops in variable lighting conditions. Second, to autonomously eliminate weeds using a robotic manipulator equipped with a laser. Third, to perform thorough controlled environment and field evaluations. This system combines robotics with laser technology, providing an efficient and eco-friendly weeding solution that boosts crop yields without the need for manual labor or chemical herbicides.

The main aim of this study is to develop an autonomous robotic manipulator equipped with a CO₂ laser as the end effector, which can move in all directions to eliminate weeds using artificial intelligence. The development of the proposed system will focus on the following objectives.

• Precisely detect weeds in onion field under variable lighting conditions.

A detection system will be developed utilizing LiDAR and high-resolution RGB camera. This system will capture detailed environmental data using sensor fusion. The LiDAR component will provide precise distance measurements and generate detailed 3D maps of the field. Concurrently, the RGB camera will capture color-based imagery, enabling the distinction between crops and weeds based on color and textural differences. The data collected from these sensors will be transmitted in real-time to the onboard NVIDIA Orin edge computer. This edge computer will feed data into a single-shot transformer-based computer vision model, specifically trained to analyze and interpret onion crops with multiple weed species. The objective of this system is to accurately detect weed plants and estimate their positions within the camera's field of view in real time.

• Development of a robotic manipulator with laser attached as the end-effector.

The robotic manipulator will be designed with a total of four degrees of freedom (DoF). It will include one prismatic joint, one revolute joint, and one spherical (ball and socket) joint. The spherical joint alone accounts for two degrees of freedom and is the attachment point for the laser. It allows for rotational movements along the X and Y axes for precise targeting. Control algorithms will be implemented to enable autonomous movement of the laser to desired coordinates. This process will involve configuring the positions of all joints, calculating inverse kinematics to determine the necessary joint angles, and planning the path of motion.

• System Integration and Evaluation

After developing the subsystems of the proposed robotic system, they will be integrated to ensure coherent and seamless functionality of the laser weeder. This integration involves harmonizing the operations of the detection system, the movement of the robotic manipulator, and the actuation of the laser system. The first goal is to minimize the latency between the real-time detection of weeds and the alignment of the laser beam on the target weed plant. The second goal is to enhance the accuracy of the system, enabling it to eliminate weeds with minimal disruption to the healthy vegetation. Once the integrated functionality of the entire system is achieved, the system will be evaluated thoroughly in controlled environment and field conditions. This would involve validating system’s performance in terms of speed and accuracy in as challenging and diverse conditions as possible.