gg009/adversarial-integrity-operations

Meta OpenEnv Drone RL (v3)

A production-grade OpenEnv environment for autonomous drone mission control with realistic 3D flight mechanics.

What It Simulates

• —3D flight physics (position/velocity/orientation updates per step)
• —Battery + budget constraints
• —Obstacle avoidance and no-fly-zone compliance
• —Sensor-mode strategy (vision, lidar, rf)
• —Adaptive hard mode with GPS spoofing and mission wave expansion
• —Mission-report submission and deterministic scoring

Endpoints

• —GET / -> game-style web UI
• —GET /web -> same as /
• —GET /health
• —POST /reset
• —POST /step
• —GET /state

Tasks

1. 1.easy_corridor_flight
2. 2.Basic corridor navigation and waypoint completion.

1. 1.medium_urban_delivery
2. 2.Denser obstacles, stronger wind, tighter no-fly margins.

1. 1.hard_adversarial_rescue
2. 2.Adaptive mission wave and GPS spoofing disturbance under strict safety constraints.

Action Space

DroneAction.action_type:

• —direct_control (throttle, pitch, roll, yaw_rate)
• —set_sensor_mode (sensor_mode)
• —set_waypoint_target (target_x, target_y, target_z)
• —ack_alert
• —emergency_land
• —submit_report

Scoring (0.0 to 1.0)

Deterministic weighted metrics include:

• —waypoint coverage
• —mission success
• —safe landing quality
• —safety (collision + no-fly penalties)
• —energy and budget efficiency
• —flight stability
• —report quality
• —required sensor-mode compliance
• —adaptive-wave completion

Baseline Inference

inference.py uses:

• —API_BASE_URL
• —MODEL_NAME
• —HF_TOKEN (or OPENAI_API_KEY fallback)
• —optional LOCAL_IMAGE_NAME

Logs are emitted in required format:

• —[START]
• —[STEP]
• —[END]

Local Run

python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
python -m uvicorn server.app:app --host 0.0.0.0 --port 7860

Open:

• —http://localhost:7860/

Validate

openenv validate

Deploy

openenv push --repo-id <username>/<space-name>