An Autonomous Spraying Robot for Sheep Houses: Hierarchical Navigation, Speed-Dependent Flow Control, and Multi-Stage Evaluation
This study presents an autonomous tracked spraying robot for sheep houses that integrates a hierarchical ROS-based navigation system with speed-dependent flow control, demonstrating high mapping accuracy, near-perfect geometric coverage, and robust obstacle handling in both simulations and physical field tests.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Inside the narrow, dusty corridors of a sheep house, the air is thick with the challenges of modern farming. The floor is often uneven or made of slats, the walls are lined with repetitive pens that look almost identical to a machine's eye, and the space is constantly shared with moving animals that can suddenly block a path. For decades, keeping these environments clean has relied on human workers walking through with sprayers, a task that is repetitive, exposes them to harsh chemicals, and often results in uneven coverage because no two people walk at the exact same speed or angle. To solve this, engineers have turned to robotics, but building a robot that can navigate such a chaotic, confined space is far more difficult than guiding a vacuum cleaner through a living room. The core challenge lies in giving the machine a reliable sense of where it is, a way to plan a route that covers every inch without getting stuck, and the ability to adjust its spraying action instantly as it speeds up or slows down.
A team of researchers at Henan Agricultural University and Zhengzhou Normal University has developed a specialized robot designed specifically for these conditions. They built a small, tracked vehicle that moves like a tank, allowing it to turn on a dime and handle uneven ground better than wheels could. This machine is equipped with a laser scanner that acts as its eyes, an inertial sensor to feel its own movement, and a computer brain running a sophisticated navigation system. The researchers did not just build the hardware; they created a three-layered software system that works together to guide the robot. The first layer decides what areas need to be cleaned, the second layer plots a smooth, safe path between those areas, and the third layer reacts instantly to obstacles, like a sheep stepping into the aisle, by slowing down or steering around it. Crucially, the system links the robot's speed directly to the flow of liquid from its pump, ensuring that whether the robot is moving slowly or quickly, the amount of disinfectant hitting the floor remains consistent.
To test if this system actually works, the team first ran thousands of simulations in a virtual sheep house. They found that by carefully tuning how the robot processed its laser scans, it could build a highly accurate map of the environment, matching the real layout with nearly 97 percent precision. When the robot was placed in a simulated environment with moving obstacles, the combined navigation system successfully completed every single test run without crashing, whereas simpler systems often got stuck or failed to find a way around. The researchers then moved to a physical mock-up of a sheep house, a real-world test area with narrow corridors and partitions. Here, the robot demonstrated its ability to follow a complex path, avoiding static obstacles and waiting patiently for moving ones to clear the way. In five separate runs, the robot covered 99.8 percent of the intended floor area, leaving almost no gaps and overlapping its path only slightly, a significant improvement over standard back-and-forth patterns.
The most critical test, however, was whether the robot could spray correctly while moving. The researchers filled the robot's tank with clean water and ran it at three different speeds: 0.08, 0.11, and 0.14 meters per second. As the robot sped up, its computer automatically increased the pump's output to compensate for the faster movement. The results showed that this speed-dependent control worked perfectly. Even though the robot was moving at different paces, the amount of water deposited on the floor remained steady at approximately 0.20 liters per square meter across all speeds. The spray pattern was also remarkably uniform, with very little variation from one side of the spray to the other. This proved that the robot could maintain a consistent cleaning dose without needing a human to manually adjust the flow or stop and start the machine.
Finally, the team took the robot into a real, occupied sheep house to see how it handled live animals. Over five supervised work cycles, the robot navigated through aisles where sheep were present. It successfully handled 55 out of 56 instances where an animal blocked its path, automatically stopping or waiting until the way was clear before continuing. The robot maintained its position with high accuracy, tracking its path with an error of less than four centimeters, and stayed connected to its map almost 100 percent of the time. While the study did not measure whether the robot actually killed bacteria or how the sheep reacted to its presence, the results confirmed that the machine is mechanically and computationally capable of performing the task. The researchers concluded that this integrated system of navigation, obstacle avoidance, and speed-controlled spraying is a viable engineering solution for the difficult environment of a sheep house, offering a reproducible way to automate a task that has long been done by hand.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.