Automation and Autonomy

Levels of Automation

There have been several levels of automation defined by researchers over the past decades. These taxonomies are platform agnostic, as automation can be applied to several areas of study. In 1999, Endsley and Kaber defined ten levels of automation as seen below.

• 1: Manual Control - Human performs all tasks.
• 2: Action Support - System assists with performance.
• 3: Batch Processing - Human selects options and systems completes tasks.
• 4: Shared Control - Human and computer generate decisions.
• 5: Decision Support - Computer generates decisions, human chooses decision. Human can also generate an option.
• 6: Blended Decision Making - Computer generates decisions and carries out with human approval.
• 7: Rigid System - Limited actions presented to human, choice fully implemented by system.
• 8: Automated Decision Making - Best option selected by computer, human can still provide input.
• 9: Supervisory Control - Best option selected by computer, human supervises.
• 10: Full Automation - System carries out all actions.

In a previous course, I discovered 5 levels of driving automation as defined by the Society of Automotive Engineers:

• Level 0: No Driving Automation.
• Level 1: Driver Assistance.
• Level 2: Partial Driving Automation.
• Level 3: Conditional Driving Automation.
• Level 4: High Driving Automation.
• Level 5: Full Driving Automation.

I found the 5 level taxonomy to be intuitive and easy to digest. There is a similar scale for UAS autonomy (autonomy is similar, but not exactly the same as automation):

• Level 0: No Automation - 100% manual control.
• Level 1: Low Automation - UAS performs some vital functions.
• Level 2: Partial Automation - UAS can control heading or altitude.
• Level 3: Conditional Automation - UAS can perform full flight but human acts as backup.
• Level 4: High Automation - UAS in full control with redundancy, human is overseeing flight.
• Level 5: Full Automation - No human input needed.

Today’s typical consumer UAS is somewhere between level 1 and 3. Some automation is in place to hold altitude and position. Level 2 is possible with some GCSs where a pre-programmed flight path is defined prior to takeoff.

Manned vs. Unmanned Operations

This question seems poorly worded, so I’ll speak to it if it was re-written: Are there different considerations for manned versus unmanned operations when it comes to automation and autonomy? Why or why not?

There are significant differences between a manned vs. unmanned system. A manned system has a much more valuable payload on board: the human crew and passengers. Automation has been built into airliners for several decades. Generally, human pilots are always considered to have control over the aircraft if the automation needs to be overridden. The two recent Boeing 737 MAX disasters are an example of a system that deferred too much control to the computer systems without enough redundancy and human override built-in. The pilots attempted to input corrective actions but were overridden by the automation system. An unmanned system, with no crew or passengers on board, still has the ability to cause injury or death to humans on the ground if a catastrophic event were to happen. The automation and autonomy applied to UAS have been implemented on smaller aircraft only. Automation or autonomy in a large, airliner or cargo jet UAS would cause destructive damage. A system with human payload will need several levels of redundancy built-in with heavy regulation from the FAA or other aviation authority.

Current State of Automation

Automation has strongly contributed to decreasing the number of air disasters in the last twenty or thirty years. The number of aviation fatalities has decreased since the 1990s as the number of total passengers has substantially increased. I am generally a proponent of automation and autonomy when engineering with several levels of redundancy. The continued advancement of automation and autonomy is something we can’t avoid, we’re just not at a point where we can hand over full control to computer systems and have complete trust. The recent issues with Boeing’s 737 MAX have signified the implications in building robust, redundant, and certified automation systems. As tragic as these events were, I expect the future of air transportation as highly automated and autonomous.