Robotics Barcelinhos

Team Barcelinhos' entry for RoboCup Junior Rescue Simulation 2026, running on the Erebus/Webots platform. The robot autonomously navigates unknown maze environments, detects and reports victims and cognitive targets (hazmat signs), avoids floor holes, and submits a 2D occupancy map to the game supervisor.

The software stack is implemented in Python 3 across three collaborating modules, an Extended Kalman Filter (EKF) for pose estimation, a log-odds occupancy grid for persistent environment mapping, and a main controller integrating reactive navigation, victim identification, and hole avoidance without any external robotics library. All 3×3 matrix operations required by the EKF are implemented by hand, keeping the codebase fully self-contained.

Navigation uses a four-level priority chain (wall stop, lateral push, steer-while-advance, heading-locked forward) evaluated every 32 ms timestep. A triple anti-stuck system with three independent detectors (GPS-based, encoder-based, and spinning detection) and an adaptive escape routine prevent the robot from being trapped in corners or cyclic patterns. Cognitive targets are identified by sampling five concentric colour rings; letter victims are distinguished by region-based dark-pixel density scores. Duplicate detections are suppressed by an 8.3 cm grid keyed to the KF position estimate. The entire controller runs under 800 lines of source code.