Defining Autonomy: What Do We Want the Aircraft to Do?

There are many ways to think about flight autonomy. In general, aviation “autonomy” means increased aircraft automation, whether the aircraft is piloted or not, and various levels of situational decision-making by aircraft systems.

At Honeywell, we often start the autonomy conversation by asking the one question that really matters: “What do we want the aircraft to do?” It’s not a philosophical question – and that is by design. This way, the answer very quickly gets us to a concrete set of requirements and functions that defines what we need to build to enable the autonomous aircraft to perform its mission. 

Different Missions Require Different Levels of Autonomy  

We have found each segment has different reasons for wanting more automation or autonomy.

In the case of advanced aerial mobility (AAM) vehicles carrying people, the primary motivation is to use precious battery power for more payload or range instead of carrying the weight and volume of a pilot. For cargo-carrying aircraft, the goal is flexibility of operations and faster turnaround time. More operations equals more deliveries, better service and more revenue.

For general aviation aircraft, it’s really about safety and simplicity. Higher levels of autonomy can make the whole task of piloting an aircraft simpler, safer, and more user-friendly, which will attract more people to learn how to fly. In business aviation, there’s the added motivation of reducing crew size and cost to address the looming pilot shortage.

Reduced crew operations is also a big motivation for airlines, which need to have a powerful and safe fallback in the event of pilot incapacitation under single-pilot operation. And, for the military, the need for autonomy covers all these things and more. Military pilots may need an autonomous wingman who can keep warfighters out of harm’s way, attack the enemy, collect intelligence and optimize survivability (for itself and for piloted aircraft).

Tracking Autonomy from Your Washing Machine to 2001’s HAL

For more than 20 years the Sheridan and Verplank¹ scale has been used as an autonomy framework. The scale provides a 10-level ranking of a computer’s involvement in making decisions:

In Level 1, the human makes all the decisions and actions – like a washing machine. In Level 2, the system offers a list of recommendations. In Level 5, the system executes the decision if the human approves. In Level 7, the system executes the decision and asks for human forgiveness later. In level 10, the system doesn’t even ask for forgiveness, and you have HAL, the independent computer with a mind of its own, from the movie “2001: A Space Odyssey.” 

In the autonomous aircraft world, we might have aircraft with varying levels of decision-making authority, autonomy and checking-in with a human pilot or supervisor – all based on the aircraft’s capabilities, what it’s carrying, its mission and the operating environment.

As we make aircraft more autonomous, we can decide to make them more or less independent of human intervention based on the speed of the decisions, the impact of the decision, the availability of communications and our ability to communicate the decision context.  For example, if an aircraft needs to make an emergency maneuver to avoid another aircraft, that might be an automatic decision and action. But if it wants to change destination, that might require human approval.

How Flight Autonomy Applies to Various Missions

Our Honeywell team believes today’s avionics can evolve organically and gracefully to add automation and autonomy elements, based on best practices for mission-critical systems that have made aviation the safest mode of transportation. Consider these use cases:

• A cargo operator could fly a fixed route in uncontrolled airspace with a highly automated, pre-programed aircraft. Such an aircraft is really the equivalent of a light rail system in the sky.  We only need five capabilities not found in avionics systems today. These include the ability to monitor aircraft system health and trigger a return to base or divert and land; automatic take-off and landing; detect and avoid; communication with the ground; and the ability to set its own transponder code to an emergency code.

• The same scenario in controlled airspace adds the need for the vehicle to interact with air traffic control and respond as required. A ground-based pilot or the aircraft itself could handle that interaction.

• The only additional capability needed to transport people in controlled or uncontrolled airspace is an ability to communicate with passengers to provide briefings and instructions. Passengers may also need to communicate with the aircraft or its supervisor.

• For an on-demand air taxi, we need to add a way to access an automated flight planning service on or off the aircraft to rapidly plan the route from the pickup point to the destination. Think about how an Uber or Lyft driver uses Google Maps on the ground. Once in the air, the passenger needs to be able to reroute to a different destination. 

• Finally, imagine an unmanned aircraft looking for multiple survivors in a search area. There are preplanned search patterns in many avionics systems today that search an area by flying a fixed pattern. An uncrewed vehicle needs to have the ability to narrow its search if a survivor is detected. If it is running low on battery power or fuel, it may need to decide to suspend the search. This kind of higher-order, goal-directed behavior and decision making is then necessary.

Clearly some technical challenges will need to be overcome to add these new capabilities to current avionics portfolios and ensure the safety of these new onboard technologies and features. Honeywell’s talented and dedicated engineering teams are making significant progress at adding new capabilities to proven avionics technologies. They are taking some cues from biology and neural science. 

Natural World Inspires Honeywell Engineers

Autonomy in natural organisms is often layered. Consider the amygdala, which is often called out as the primordial human brain. The amygdala is wired for fast-firing responses like fear, anxiety and fight or flight and is the source of the human drive for self-preservation.

The autonomous aircraft equivalent of the amygdala are things like primary flight controls or detect and avoid systems. These are the high bandwidth, safety-critical systems responsible for keeping aircraft safe and avoiding collisions.

The neocortex is the outer layer of the brain that handles sensor fusion, motion planning, abstract reasoning, navigation and language. The autonomous aircraft equivalent of the neocortex is the autonomy executive software that can select plans or formulate them online in response to pre-planned or dynamic conditions like weather, air traffic and service demands.

This leads us to some useful engineering questions involving hierarchical autonomous systems, optimization, integration, consensus and synchronization of data and decisions. The ability to think about autonomy from different perspectives ensures a more robust solution.