AERESCUE

The controlled, computational District-level prototype is complete. SPIKE Prime movement testing is the next planned stage; drone integration and real rescue operations are long-term goals.

Overview

AERESCUE (Assistive Emergency Rescue and Support System) is a shore-based, dual-camera computer-vision research prototype for simulated water-rescue scenarios. An operator manually marks a rescue target in both camera views while the system tracks an orange-folder surrogate as the moving object. Triangulation estimates their positions and relative distance, then the console presents guidance such as LEFT, RIGHT, FORWARD, and STOP/HOLD.

The District-level prototype demonstrated dual-camera target localization, moving-object detection and tracking, relative-distance estimation, and basic directional guidance. It supports trained operators in controlled simulations; it does not autonomously conduct physical rescues.

Congressional-level work improved camera calibration and tracking and reacquisition reliability. The next planned step is to send guidance commands to a LEGO SPIKE Prime drive base used as a controlled moving-platform surrogate and evaluate how reliably the commands produce movement.

The Division-level roadmap includes continuous PD-based guidance, measuring path error and response latency, stop/hold behavior during tracking or communication failures, manual override and emergency stop, position holding near the selected target, and recording data for closed-loop evaluation. Drone integration and flotation-device delivery remain long-term goals.

My Role

Developer of the AERESCUE research prototype.

Structure

• cameras — Two fixed shore-based cameras observe the manually selected target and the moving-object surrogate.
• tracker — OpenCV trackers detect and follow the orange-folder surrogate, then report observations for each camera.
• server — The Flask server manages camera and tracker processes, receives observations, and keeps the latest tracking state.
• console — The browser console triangulates positions, estimates relative distance, and displays directional guidance.

Technologies

Results & limitations

• Designed and evaluated for controlled simulated rescue scenarios, not real-world rescue operations.
• The rescue target is selected manually by an operator.
• The current prototype generates computational guidance and does not autonomously control a drone or other vehicle.
• Continuous PD control, fail-safe behavior, and physical platform integration remain planned work.