Kavosh_Simul

Kavosh_Simul is an autonomous Rescue Simulation robot designed around efficient wall inspection, reliable token detection, and modular software integration. The robot uses a two-wheel structure with two side-facing cameras, GPS/location sensing, IMU, LiDAR, color sensing, and distance sensors, including a downward-facing distance sensor for early hole detection. Its software is organized into independent modules for sensing, vision, mapping support, navigation, motion control, and reporting.

The main strength of our system is its perception pipeline. Instead of sending every full camera frame directly to a neural network, the robot first searches for meaningful wall candidates, crops only the region of interest, and then sends the ROI to a YOLO11s detector for Greek victim symbols. Cognitive targets are handled separately using an OpenCV-based rule pipeline with HSV segmentation, circular filtering, ring sampling, and numerical classification. To improve reliability, detections are verified through confidence checks, temporal confirmation, 3D fake-victim rejection using depth information, and GPS/yaw/LiDAR-based duplicate suppression.

What sets our robot apart is its focus on practical reliability rather than raw detection count. Our next development step is to complete our Python-based SLAM mapping system and connect it with wall-focused path planning to inspect all important wall areas with minimum unnecessary travel.