7-Inch Autonomous AI Guard Drone Build

KnowlEdge2 AI's flagship autonomous drone. Jetson Orin Nano, ArduPilot, RTK GPS, auto-landing, YOLOv8 on-board. Full hardware guide.

Updated March 2026

6S / 24V
Power
8–12 min
Patrol time
67 TOPS
On-board AI
ArduPilot 4.6
Firmware

🛡️ What This Drone Does

This is the KnowlEdge2 AI guard drone — autonomous AI-powered property patrol. The Jetson Orin Nano Super runs Isaac ROS Visual SLAM for navigation and YOLOv8n TensorRT (8.8MB FP16) for real-time object detection. MAVROS2 bridges ArduPilot to ROS2.

The ZED-F9P RTK GPS provides centimeter-level landing accuracy on the charging pad. The pilot holds RC only as a safety override — the drone arms, takes off, patrols, detects, and lands autonomously.

One Click

Add complete build to Amazon cart

All 8 parts from this guide added at once. Review quantities before checkout — motors come in 4-packs; buy 2 packs for 5 total (4 + 1 spare).

Add All to Cart →

Hardware Stack

We Fly This
Matek Systems H743-Slim V3 Flight Controller
View on Amazon →

Matek Systems H743-Slim V3 Flight Controller

4.6/5

Our flight controller. STM32H743, 8 UARTs, 30×30mm. Runs ArduPilot Copter 4.6.2. Enough serial ports for GPS + companion + RC + telemetry simultaneously.

Check Price on Amazon
Edge AI Core
NVIDIA Jetson Orin Nano Super Developer Kit (8GB)
View on Amazon →

NVIDIA Jetson Orin Nano Super Developer Kit (8GB)

4.7/5

The brain. 67 TOPS AI inference, JetPack 6.2, CUDA 12.6, TensorRT 10.3. Runs YOLOv8 real-time, ROS2 Humble, Isaac ROS SLAM. The autonomy engine.

Check Price on Amazon
Our GPS
SparkFun GPS-RTK2 Board (ZED-F9P)
View on Amazon →

SparkFun GPS-RTK2 Board (ZED-F9P)

4.6/5

RTK GPS with centimeter-level accuracy. UART to FC SERIAL3. ArduPilot GPS_TYPE=17. Essential for precision auto-landing on charging pad.

Check Price on Amazon
DJI O3 Air Unit
View on Amazon →

DJI O3 Air Unit

4.5/5

1080p/60fps digital FPV, ultra-low latency, on-board storage. Used for operator monitoring. Pairs with DJI Goggles 2.

Check Price on Amazon
We Use These
iFlight Xing2 2809 1250KV Motors (4-Pack)
View on Amazon →

iFlight Xing2 2809 1250KV Motors (4-Pack)

4.5/5

Our motors. Cinelifter-grade for heavy payload (Jetson). Buy 5 total — one spare is mandatory on a build this complex. 6S optimized.

Check Price on Amazon
We Use This
Foxeer Reaper F4 65A 4-in-1 ESC (AM32)
View on Amazon →

Foxeer Reaper F4 65A 4-in-1 ESC (AM32)

4.4/5

65A continuous on 3-8S, 30×30mm. Configure via esc-configurator.com (not BLHeli Suite 32). No BEC — power GPS/receiver from FC 4V5 pad.

Check Price on Amazon
We Use This
RadioMaster RP3 ELRS 2.4GHz Receiver
View on Amazon →

RadioMaster RP3 ELRS 2.4GHz Receiver

4.7/5

The exact receiver on our build. Diversity nano, UFL antennas, T6/R6 pads. SERIAL7_PROTOCOL=23, RC_PROTOCOLS=512. Confirmed working.

Check Price on Amazon
Tattu R-Line V4.0 6S 1300mAh 130C (2-Pack)
View on Amazon →

Tattu R-Line V4.0 6S 1300mAh 130C (2-Pack)

4.6/5

High-C 6S packs for the 7" + Jetson payload. ~8-12 min patrol time per pack. Buy two minimum for meaningful patrol sessions.

Check Price on Amazon

Core Components at a Glance

7-Inch Guard Drone Core Stack
Product Buy
Our Choice Jetson Orin Nano Super (8GB)
Amazon ↗
Our FC Matek H743-Slim V3
Amazon ↗
Our GPS SparkFun ZED-F9P GPS
Amazon ↗
DJI O3 Air Unit
Amazon ↗

Software Stack (All Open Source)

Flight Controller ArduPilot Copter 4.6.2
Companion OS JetPack 6.2 (Ubuntu 22.04)
Middleware ROS2 Humble + MAVROS2
AI Vision YOLOv8n TRT FP16 (8.8MB engine)
SLAM Isaac ROS Visual SLAM 3.2.6
Simulator Isaac Sim 4.5 + Pegasus 5.1

Confirmed Test Results

  • Isaac Sim + Pegasus SITL: Full autonomous mission PASSED — arm → 5m → 5-waypoint box → land (2026-03-19)
  • All 4 motors spinning at 30% throttle confirmed (2026-03-20)
  • YOLOv8n TRT engine: 8.8MB FP16, real-time on Jetson Orin Nano Super
  • Battery voltage: ~22.6V on 6S confirmed via MAVLink

AI payload changes your power budget: The Jetson Orin Nano Super draws ~10W at idle and up to 25W under full YOLOv8 + SLAM inference. At 6S (22V), that's an extra 1.1A of continuous draw your battery sizing estimate must include. On a 1300mAh pack averaging 35A total, AI inference alone cuts flight time by roughly 8–12%. If patrol time matters, disable unused Isaac ROS nodes between detection events to reduce idle draw.

MAVLink 2 is required: The Jetson companion connects via SERIAL1. Set SERIAL1_BAUD=921 (921600) and SERIAL1_PROTOCOL=2 (MAVLink 2). MAVLink 1 drops critical message types needed by MAVROS2. Mismatch here is the most common reason ROS2 topics appear empty on first boot.

Affiliate disclosure: As an Amazon Associate, KnowlEdge2 AI earns from qualifying purchases. Links on this page may be affiliate links. Full disclosure →