Helix Command
The decision layer for rescue robots: verified detections, a plain-code safety gate, human-approved lifts. Simulation prototype.
Built at AgentX Hackathon 2026
Team qyrixx: Hemamalini Yuvaraj, SANJAY GANESAN, Roumaan Afridi_ROBO, EZHILAVAN G M
Built with Node.js, HTML canvas, WebRTC (PeerJS), URDF, Gazebo, ROS 2, Meta Muse API, Vercel, GitHub
About the project
Problem: rescuers die entering fires, floods, collapses and wells, and a rescue robot needs a decision layer people can trust.
Solution: Helix Command sits between a robot's planner and its actuators. The operator taps DEPLOY on a phone; 1 to 3 units search with heat camera, microphone, CO2 and radar, and a casualty counts only when two different sensor types agree. A deterministic safety gate written as plain code (no AI model) checks every action: weight, heat, water depth, structure. A human approves lifts, and every decision goes into a tamper-evident SHA-256 log.
Four scenarios: building fire (90 kg solo lift, 170 kg pair lift, re-planning, dogs), flood (swimming with flotation, towing to shore, trickle-charging from the current), collapse (listening for voices, clearing blocks under a stability check), and well rescue (winch, gas check and ventilation, tethered rescue pod for children). Units hand over casualties on low battery, and a flashing alarm names the human help needed (lifting crew, fire crew, divers, crane or rope team, medic).
Impact (in simulation): across 100 randomised runs per scenario the gate allowed 0 safety violations, versus 664, 681 and 1,158 without it. 72 automated tests pass.
Honest scope: this is a simulation prototype. Sensors, camera frames and the robot are simulated, limits such as 90 kg are design targets, and no physical robot exists yet. The software is finished and the design is hardware-ready.