While the YOLO-based pest detection shows modern promise, the "autonomous" label is stretched thin by a navigation system that still relies on colored tape. It is a practical academic exercise that succeeds in integration but lacks the sophisticated localization required for real-world scalability.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
26008 – LYCO TOMIAdded:
Hello. Today we'll be showcasing the development of our greenhouse companion Licotomi. My name is Estellan Aragon.
I'm a mechanical engineer, the team lead, and I was in charge of the fabrication of our robot.
My name is Tal Monet, and I'm a computer engineer. I was in charge of the electrical fabrication and the programming for autonomous navigation.
>> My name is Kayla Melindorf. I'm a bios engineer, the procurement lead, and I was in charge of project management.
>> Hello, my name is Roberto Enrique Diaz.
I am a bios systems engineer and I was in charge of the pest identification model.
>> My name is Ramon Partila. I am a software engineer and I was in charge of the mobile application software.
>> My name is Walter Davenport. I'm a mechanical engineer and I was responsible for the design and fabrication of the robot.
The engineering design process is an important part of learning to be an engineer. In this poster and as you'll experience through this video, we face the problem presented to us. We do our research and with input from our sponsor determine the required verifications that must be achieved to have a success.
With these constraints in mind, we follow through during the design process, ensuring we remain within the budget and the scope of the project. The results showcase how we've managed throughout this course and give us valuable experience towards becoming engineers. Our project liotomy short for lightweight yield crop optimizer tomato interface aims to reduce manual labor for staff at the controlled agriculture research greenhouse and eliminate the risk of allergens that arise from the deployment of pollinating bees. The research greenhouse also implements long working hours in inspecting each sticky trap in the greenhouse looking for insects harmful to the plants which has slowed operations and increased labor costs.
Our initial research for this project started by learning about tomato crops which can grow up to 13 ft tall.
Additionally, we determined that pollination can be achieved through agitating tomato flowers. To determine the best design, our trade studies compared different pollination methods and the mechanisms needed to reach all tomato flowers. The pollination methods we considered were vibrating the stems, stimulating the guide wires, and using fans to agitate the crops. The mechanisms we considered were a scissor lift, a cascade elevator, and a static structure.
Our initial research for the PES identification model focus on understanding the object detection and machine learning, which led us to select the YOLO B5 model due to its fast and accurate performance for real-time computer vision tasks. After establishing the model, we evaluated hardware options and incorporated the Oaklight camera based on a sponsor request. Through documentation and existing implementations, we confirmed that YOLO V5 could be successfully deployed with the Oaklight camera using the depth AI framework, allowing for efficient image capture and real-time best detection.
Our project system requirements were established with input from the sponsor and the scope of the project. These are separated by the respective subsystems, mobilization, vertical, control, and power systems. We'll go over some of the requirements we've established as key to our project design.
System requirement 1.1. The Tomy system shall autonomously travel through predefined greenhouse rows. As per standard RAI R15, August 1st, 2020, line following robots stop when obstacles interrupt the line. This standard ensures that there are no collisions with people or crops.
System requirement 1.5. The Tomy system shall be able to traverse over obstacles of a maximum of 2 in in diameter. The greenhouse Tommy will be deployed in has PVC pipes all under 2 in in diameter along its navigation route. System requirement 2.1. The Tomy system shall pollinate tomatoes through agitation.
Based on the scope of our project, Tommy must be capable of pollinating the tomato crops. Therefore, any significant agitation of the tomato flowers will achieve successful fertilization.
>> Here is our model design in Solid Works.
This shows the full integrated system, our vertical system and mobilization system. Additionally, the model showcases the trade study results, the cascade elevator and fans for crop agitation.
Our first critical design element was a vertical lift system. Based on the design, the system was made by welding aluminum U channels in a nested configuration while movement was supported by roller bearings. The finished fabrication shows the aluminum cascade elevator equipped with two axial DC fans utilized to agitate the tomato flowers.
Fabrication for our second critical design element started when fabrication for a vertical lift system concluded.
This entailed incorporating both the fabrication of the rover and the development of the electrical system that would give our robot movement. The mobilization system depicted here is comprised of a steel welded frame with four 10-in pneumatic wheels.
The pest identification model was developed by taking images of sticky traps at the research greenhouse. The insects within these images were classified manually through Roboflow from which we could build a database for the YOLO B5 to use to identify the pests. This data set was then exported and inputed into a working Python model that uses the Oaklight camera to detect these pests.
The electrical system is designed to meet the specific requirements of our critical design elements. One of the key components is the LAR which measures and helps adjust the height of the vertical lift system. This is important for safety to ensure the lift does not overextend while also providing exact height measurements for the tomato plants. Another critical component is the Raspberry Pi which processes our control algorithms and reads the camera feeds allowing the system to successfully navigate the greenhouse and identify pests. Finally, the system is powered by a 12volt lithium battery and routed through a fuse block to safely protect all the electrical components.
The software system is designed to have multiple subsystems operating concurrently. The main software contains the computer vision pipeline for pest identification, computer vision pipeline for obstacle detection, SQL database of scene pass as well as the web server to host the backend for the mobile Android app acting as a user interface. The mobile app is the main interface for the system allowing the user to start the mission drive the system manually and view and receive notifications from the database. The script building all the subsystems was made using Python and various computer vision libraries. The web server is built through Flask and the mobile app is built for Android through Java and Android Jetack.
>> From these videos, we can see the finished results of Leico Tell me. We can see that it successfully navigates through greenhouse rows both autonomously and by manual control through the app.
The robot's autonomous navigation works by creating a continuous feedback loop between its cameras and its motors.
First, the software processes the active camera feed by isolating the specific tape color. Red for the straightaways and green for the turns. The code calculates the center of the line and the exact angle to create a steering path for the robot to follow. When the camera loses sight of the tape at the end of the row, the built-in state machine takes over, which automatically switches to the opposite camera and adjusts the vertical lift position to ensure proper pollination.
>> After building the data set with the annotated images, we custom trained a YOLO V5 object detection model. This model is exported to the pest identification computer vision pipeline in the main software. When the user starts the mission, the system is actively searching for pests and records each observed pest into a database. For pests identified as high priority, the software system sends locations to the user through the mobile app.
Throughout this project, the team has learned many new skills that we will be taking with us into our future endeavors. soft skills like time management, conflict resolution, and teamwork, as well as hard and technical skills like welding, technical writing, and programming. Overall, the Tomy team has learned a lot about the design process and how time and project management is key to a successful and fulfilling final result.
From all of us at Team Tomy, I hope you had as much fun learning about our project as we did making it. Thank you for listening.
Related Videos
Beyond Robotics | European Rover Challenge 2026
beyondrobotics
189 views•2026-06-01
Beatbot Sora70: JetPulse Technology and AI obstacle avoidance and navigation!
DroidModderX
26K views•2026-06-02
Tesla FSD 14.3.3 Hits Phoenix Streets - FIRST LOOK
anthonystesla
114 views•2026-05-29
Elon Musk Just Revealed Fremont Line for Optimus Gen 3 Mass Production
TheAINexusOfficial
180 views•2026-05-30
人機一体「零式人機 ver.2」 子ども企画【おもしろ発見!モビリティー】 #乗り物 #automobile #robot #shorts
KyodoNews
1K views•2026-05-28
China’s New Luna AI Robot Looks Shockingly Human...
NextGenHumanoids
850 views•2026-05-28
Reachy Mini: the $300 open source robot you can actually hack — Andres Marafioti, Hugging Face
aiDotEngineer
662 views•2026-05-29
柔軟指×AI画像処理食品の仕分け作業システム!#柔軟指 #ロボット #自動化 #製造業をもっと盛り上げたい
KiQ_Robotics_Corp.
113 views•2026-05-28











