Documents

Running Slides – our Progress

MRSD Capstone — Running Slides

Reports

• Conceptual Design Review
• Spring Test Plan
• Critical Design Review Report
• Fall Test Plan
• Final Report

Presentations

• Preliminary Design Review
• Critical Design Review
• Spring Validation Demo
• Standards and Regulations

Progress Reviews – Spring Semester

• Progress Review 1
• Progress Review 2
• Progress Review 3
• Progress Review 4

Progress Reviews – Fall Semester

• Progress Review 7
• Progress Review 8
• Progress Review 9
• Progress Review 10

Individual Lab Reports (ILR)

• CAD Design Files

Component Testing & Experiment Results

Pot Detection Model Unit-Test

We trained two YOLO models for pot detection, YOLO11n and YOLO11s, which differ in the number of model parameters. We evaluated the model prediction accuracy on a test set consisting of 97 images containing 359 pot instances.

As shown in the table below, both models achieve strong performance on the evaluation set. To further reduce false positives during deployment, we apply a confidence threshold, which may slightly increase missed detections but improves overall reliability.

Pot Detection Model Accuracy

Metric	YOLO11n Result	YOLO11s Result
mAP@50	0.955	0.956
mAP@50-95	0.877	0.881
Precision	0.940	0.944
Recall	0.922	0.931

GPS Localization Static Open-Sky Unit-Test

We evaluated the GNSS-RTK localization subsystem under static open-sky conditions to measure position stability. The robot remained stationary while GPS position data was collected and converted into local Cartesian coordinates. This test evaluates localization repeatability and sensor jitter, rather than absolute global position accuracy.

As shown in the table below, the GNSS-RTK module produced highly stable position estimates. The total 2D position stayed within 0.73 cm of the average position, and the RMS variation was 0.28 cm. These results show that the localization subsystem provides stable input for autonomous navigation and planting location estimation.

GPS Localization Static Open-Sky Unit-Test Results

Metric	X (cm)	Y (cm)	Total 2D (cm)
Max horizontal drift	0.63	0.56	0.73
Total position range	1.25	1.10	1.66
RMS variation	0.23	0.16	0.28

The low drift and RMS variation indicate that the GNSS-RTK subsystem is stable enough to support the project requirement of identifying planting locations within 1 m precision.

Design Brainstorming