OGEE
Document register
- Poster1 pagePublished
- Presentation videoYouTubePublished
- Bill of materials2 pagesPublished
- Team description paperNot shared
- Engineering journalNot shared
- Source code10 KB · GitHubPublished
Sharing each document is the team's decision. “Not shared” means this team chose not to publish it, or did not submit one — not that it is missing from the archive.
In their words
This paper presents the autonomous virtual rescue robot developed by Team OGEE for the RoboCupJunior Rescue Simulation 2026 competition.The system is designed to navigate a highly hazardous, simulated rescue environment completely autonomously, identifying victims and generating real-time environmental maps. Our robot platform integrates a differential-drive chassis with a specialized sensor array featuring high-resolution LiDAR, multi-angle RGB cameras, a Global Positioning System (GPS), and an Inertial Measurement Unit (IMU), staying strictly within the 3000-unit competition budget constraint. What sets our system apart from competitors is its highly integrated, modular sensor-fusion architecture. Locomotion is governed by an innovative dual-Dijkstra pathfinding system equipped with path-smoothing turn penalties and a strict hysteresis gate to eliminate frontier oscillations in tight corridors. For victim identification, the robot utilizes a custom OpenCV computer vision pipeline featuring a geometric Kasa least-squares circle-fit filter to discard fake 3D distractors and a robust connected-components apportionment engine to decode complex, multi-layered cognitive targets under severe camera skew. Simultaneously, a grid-based SLAM module fuses spatial sensor inputs into a 5 mm resolution matrix to secure the mapping bonus multiplier. The resulting architecture balances computational efficiency with exceptional resilience to sensor noise.
Poster
Read the text of this document — 367 words
OGEE
Team Members: Ege Serter, Osman Baran Ayaydın, Göktuğ Aslanoğlu, Erkam Tuna Bayoğlu
Team Mentor: Gökhan Doğan
Past Experience Sensor Strategy Algorithms and Methods
Erkam Tuna Bayoğlu: TÜBİTAK 2204-A Tech Design 1. LiDAR: Mounted on the top center. Used for1. Dynamic Navigation (Dijkstra)
2nd Place (Turkey) & 2024 FRC Marmara Regional complex obstacle detection (especially in non- Exploration Planner: Applies a "Turn Penalty" to
Winner. grid Area 4) and high-resolution (360-degree) prevent zig-zagging and ensure smoother
Ege Serter: CS & Algorithm expert. Gold Medalist mapping. trajectories.
(Istanbul Science Olympics) & 2x Bronze Medalist Targeted Planner: Employs a Hysteresis Gate to
(TÜBİTAK National Science Olympics). 2. 3x RGB Cameras: Side-mounted cameras prevent rotational oscillation in narrow
Osman Baran Ayaydın: Software specialist. 1st & 2nd simultaneously scan for victims in narrow corridors.
Place Winner (Turkey) in the TÜBİTAK 2204-A corridors, while the downward-facing front 2. Advanced Computer Vision (OpenCV)
Research Projects Competition. camera handles floor analysis (time-consuming Uses custom HSV filtering to reject wall reflections
Göktuğ Aslanoğlu: Computer Vision & Autonomous brown swamps, black holes, and silver and fake 3D targets.
Systems expert. TÜBİTAK BİLGEM & UME intern; checkpoints). Cognitive Targets: Analyzes ring thickness via
experienced in underwater CV modules and LLMs. 3. GPS and IMU Integration: Absolute position connected-components. Partially merged rings
(GPS) and rotational heading (IMU Gyroscope) are mathematically extrapolated to 5 based on
data are fused to ensure flawless SLAM thickness ratios.
mapping. Letter Victims: Accurately identifies letters by
counting white components
LOOP while simulation runs:
3. Grid-Based SLAM Mapping
Read Sensors (LiDAR, GPS, IMU, Cameras) Safety Margin (Inflation): Generates a 5mm
IF tilt > 0.01: SKIP iteration (Prevent map error) resolution LiDAR grid. Walls are "inflated" by
Update Map (Pose, Visited FOV, Obstacles, Holes) 2.5cm and holes by 1.2cm to guarantee
SWITCH robot_state:
collision-free paths.
0 (Explore): FOV Tracking: Safe areas within the camera's
Follow Dijkstra exploration path. Field of View (FOV) are instantly marked as
IF target detected -> State = 1 "visited" using OpenCV fillPoly. A tilt-guard
1 (Approach Target):
Navigate to 0.06m from target.
prevents map corruption on uneven terrain.
IF reached -> State = 2
2 (Align & Identify):
Rotate to face target.
IF aligned -> Run Vision algorithms, Mark Done
1 page, rendered as images so they load quickly. The text above is the document's own, extracted from the PDF.
Presentation video
Hosted on YouTube. The player loads only when you press play.
Bill of materials
Read the text of this document — 107 words
Team name: OGEE
# Name Software's Tool/Library Description
1 Git Software's Tool Version control for managing functions
2 Open CV Library Image processing library
6 Matplotlib Library Python graphics generation library
4 Visual Studio Code Software's Tool Software
5 Numpy Library Array and math library
Local Currency Local Currency U.S.A. Dollars
Turkish Lira R$0,00 $0,00
Cost (Local Cost (USA
Source Autor Currency) Dollars)
Software Freedom
Git - Descargas Conservancy FREE FREE
OpenCV - Biblioteca abierta de Computer Vision Opencv.org FREE FREE
Matplotlib — Visualización con Python Matplot lib dev team FREE FREE
Visual Studio Code: edición de código. Redefinido Microsoft FREE FREE
NumPy Numpy developers FREE FREE
2 pages, rendered as images so they load quickly. The text above is the document's own, extracted from the PDF.
Source code
The team's own source code, 10 KB. It is a download rather than part of this page, because a zip is something you open on your computer. It comes from GitHub, which some school networks block.