TF-Luna applications
How TF-Luna is used in our indoor drone stack.
Semantic search over the full team-docs corpus — reference, journal, gallery and dev-log. Powered by a local Qdrant vector index built from Xenova/all-MiniLM-L6-v2 embeddings.
How TF-Luna is used in our indoor drone stack.
TF-Luna technical specs: 8 m range, 100 Hz, ±6 cm accuracy, 2° FOV.
Driver overview for the TF-Luna sensor; for the ESP32-S3 implementation see the firmware-specific page.
Which of EXP-7's input sensors actually drive navigation: we fed the model noisy inputs one at a time and watched coverage.
Firmware-agent implementation notes: frame sync via scan-for-magic 0x59 0x59, UART buffer 512, health flag (amp<100 / 0xFFFF), 7 host unit tests.
Transferring a TF-Luna-trained RL policy to the real sensor: which sigma noise to add to Gazebo's gpu_lidar, domain-randomization setup, and critical ranges.
Cloning iris_with_ardupilot, adding TF-Luna, indoor.parm, propeller colors, IMU + NavSat plugins.
DDS RMW implementations, QoS profiles, transport overhead for a drone on ROS 2 Jazzy. Latency numbers across sensor stream and command configurations.
Drone servos and actuators: SG90 micro-servo for sensor sweep, motor ESCs (in power_management).
Tower Pro SG90 — 9-gram micro-servo 0°-180° for the sensor arm. PWM 50 Hz, 4.8-6 V supply. Used for the TF-Luna sweep on iris_claudedrone.
SG90 angular accuracy for sensor sweeps and sweep-data storage. Settle time 50-120 ms/step for TF-Luna 100 Hz; storage via RAM ring buffer flushed at scan end.
Sensor harness firmware: ESP32-S3 + Arduino / ESP-IDF, lidar/ToF/optical-flow drivers, host-testable C++ patterns.
Short-to-medium-range Time-of-Flight sensors: VL53L0X (2 m), VL53L1X (4 m). I2C interface, multi-sensor arrays.
Driver pattern for TFMini-S — similar to TF-Luna, with frame-layout differences.
How our best and previous models survive real sensor noise across 30 no-retrain experiments — and why one forward sensor is mission-critical.
The drone learns to feel inertia, tilt forward to fly, wobble like a real quadcopter, and aim its TF-Luna fan-scan on purpose.
1D rangefinder readings as policy input — TF-Luna sweep, VL53L0X array.
Mount geometry and position of each sensor (lidar, IMU, optical flow).
How a drone learned to actively scan its surroundings with a cheap rangefinder before flying, making active perception a real skill.
┌─────────────────────────────┐
┝──── you've got mail ───────▮
└─────────────────────────────┘╭──── support_the_build.sh ───╮ │ components · sim · bench │ ╰─────────────────────────────╯