Fatskills
Practice. Master. Repeat.
Study Guide: AI and Autonomous Systems: Drones and unmanned systems
Source: https://www.fatskills.com/ai-for-work/chapter/ai-autonomous-systems-drones-and-unmanned-systems

AI and Autonomous Systems: Drones and unmanned systems

By Fatskills Exam Guides Team — the exam nerds behind 28,500+ quizzes and 2.1M practice questions across 500+ global exams.

⏱️ ~5 min read

Drones and Unmanned Systems: A Practical Study Guide for Autonomous Systems

What This Is

Drones and unmanned systems are autonomous or remotely operated vehicles (aerial, ground, or maritime) that perform tasks without a human onboard. They matter in everyday work because they reduce risk, cut costs, and enable scalable operations—from inspecting infrastructure to delivering medical supplies. Example: A utility company uses drones to autonomously inspect power lines, detecting faults 10x faster than manual crews while eliminating worker falls.


Key Facts & Principles

  • Autonomy Levels (0–5): A scale measuring how much a system operates without human input. Level 0 = full manual control; Level 5 = fully autonomous (no human oversight). Example: A drone that follows a pre-programmed flight path (Level 2) vs. one that dynamically avoids obstacles in real time (Level 4).

  • Sensor Fusion: Combining data from multiple sensors (e.g., cameras, LiDAR, GPS) to improve accuracy. Example: A drone uses LiDAR for terrain mapping and thermal cameras to detect heat leaks in buildings.

  • Geofencing: Virtual boundaries that restrict a drone’s movement to a predefined area. Example: A warehouse drone is programmed to stay within a 500m radius of the loading dock to avoid airspace violations.

  • Swarm Intelligence: Multiple drones coordinating as a group to complete tasks (e.g., search-and-rescue, agriculture). Example: 20 drones map a wildfire’s perimeter in parallel, sharing data to update the fire’s spread in real time.

  • Regulatory Compliance (FAA, EASA, etc.): Rules governing drone operations, such as Part 107 (U.S. commercial use) or Open Category (EU). Example: A drone must stay below 400 ft and avoid flying over people without a waiver.

  • Edge Computing: Processing data on the drone itself (rather than sending it to the cloud) to reduce latency. Example: A drone inspecting a wind turbine analyzes cracks in real time using onboard AI, instead of uploading video for later review.

  • Fail-Safes: Redundant systems to prevent crashes (e.g., return-to-home on signal loss, parachutes). Example: If a drone’s battery drops below 20%, it automatically lands at its takeoff point.

  • Payload vs. Endurance Trade-off: Heavier payloads (e.g., high-res cameras) reduce flight time. Example: A drone with a 30-minute battery life can fly for 20 minutes with a LiDAR sensor but 28 minutes with just a GoPro.


Step-by-Step Application

  1. Define the Mission:
  2. Ask: What problem are we solving? (e.g., "Reduce pipeline inspection time by 50%.")
  3. Match the drone’s capabilities (sensors, autonomy level) to the task. Example: Use a thermal camera for gas leaks, not a standard RGB camera.

  4. Select the Right System:

  5. Choose between fixed-wing (long endurance, large areas) or multirotor (precision, VTOL).
  6. Example: Fixed-wing for mapping a 100-acre farm; multirotor for inspecting a single cell tower.

  7. Plan the Workflow:

  8. Pre-flight: Check weather, geofencing, and battery levels.
  9. In-flight: Monitor telemetry (altitude, speed, signal strength).
  10. Post-flight: Process data (e.g., stitch drone imagery into a 3D model using software like Pix4D).

  11. Ensure Compliance:

  12. Register the drone (if >250g in most countries).
  13. Obtain waivers for night operations or flying over people.
  14. Example: In the U.S., file a LAANC request for controlled airspace.

  15. Test and Iterate:

  16. Start with a manual flight to validate the mission plan.
  17. Gradually increase autonomy (e.g., waypoint navigation-obstacle avoidance).
  18. Example: First fly a drone manually over a construction site, then switch to autonomous mode once the route is proven safe.

  19. Integrate with Existing Systems:

  20. Connect drone data to workflows (e.g., auto-generate reports in Esri ArcGIS or DroneDeploy).
  21. Example: A solar farm’s drone inspection data feeds into a maintenance ticketing system.

Common Mistakes

  • Mistake: Ignoring battery life limits. Correction: Always plan for 20% buffer (e.g., if a drone’s max flight time is 30 minutes, assume 24 minutes of usable time). Why: Wind, payload, and temperature affect battery drain.

  • Mistake: Overlooking line-of-sight (LOS) requirements. Correction: In most countries, drones must stay within visual line of sight (VLOS) unless you have a waiver. Why: Beyond VLOS (BVLOS) operations require additional certifications and safety measures.

  • Mistake: Assuming all drones are "autonomous." Correction: Clarify the autonomy level (e.g., "This drone can follow waypoints but can’t avoid obstacles"). Why: Mislabeling leads to unrealistic expectations and safety risks.

  • Mistake: Skipping pre-flight checks. Correction: Use a checklist (e.g., propellers secure, GPS lock, firmware updated). Why: 60% of drone incidents are caused by human error (FAA data).

  • Mistake: Not accounting for data processing time. Correction: Budget time for post-flight analysis (e.g., 1 hour of flight = 3–5 hours of data processing). Why: Raw drone data (e.g., 4K video, LiDAR scans) requires significant computing power to process.


Practical Tips

  • Use "Dull, Dirty, Dangerous" as a Filter: Prioritize drone use for tasks that are repetitive, hazardous, or inaccessible to humans. Example: Inspecting a bridge for corrosion vs. filming a wedding (where human creativity is valued).

  • Leverage Off-the-Shelf Autonomy: Start with pre-built autonomy stacks (e.g., PX4, ArduPilot) before custom-coding. Example: Use DroneDeploy’s automated flight app instead of building your own.

  • Plan for Data Overload: Drones generate terabytes of data quickly. Use cloud storage with AI tagging (e.g., AWS S3 + Rekognition) to organize and search footage.

  • Train for "Lost Link" Scenarios: Simulate signal loss during training to practice fail-safes (e.g., return-to-home). Example: Fly in an open field and deliberately turn off the controller to test the drone’s response.


Quick Practice Scenario

Scenario: Your team is using drones to inspect a 50-mile oil pipeline. The drone keeps losing GPS signal in valleys, causing it to drift off course. Question: What’s the most practical fix?

Answer: Switch to visual-inertial odometry (VIO) or LiDAR-based navigation for areas with poor GPS. Explanation: VIO uses cameras and IMUs to estimate position when GPS is unreliable.


Last-Minute Cram Sheet

  1. Autonomy Level 4 = High automation (human override possible); Level 5 = Fully autonomous.
  2. LiDAR = Best for 3D mapping; thermal = Best for heat detection.
  3. Geofencing = Virtual boundaries to keep drones in/out of zones.
  4. Part 107 (U.S.) = Commercial drone license; Open Category (EU) = No license for <25kg drones.
  5. Edge computing = Process data on the drone to reduce latency.
  6. Swarm = Multiple drones coordinating (e.g., search-and-rescue).
  7. Fail-safe = Redundant systems (e.g., return-to-home on signal loss).
  8. Payload vs. endurance = Heavier sensors = shorter flight time.
  9. VLOS-BVLOS – Beyond visual line of sight requires waivers.
  10. Battery buffer = Always plan for 20% less flight time than advertised.