Motivation for developing the theory

It is inherently hard to keep aircraft properly separated using conventional technology. This is one reason why the average U.S. air traffic controller currently makes about $170,000 a year.

At the root of the problem is the old-fashioned, veridical display of traffic information. Veridical displays are based on a representation of physical reality. They are maps. And the root problem (in our opinion) is that maps are not necessarily the best way to approach aircraft conflict resolution because they require the user to solve both the conflict detection and the conflict solution problems, both of which are hard problems.

Naturally, there have been serious attempts to solve conflict detection and resolution with technology. Right now, the world's best veridical CDTIs are arguably the NASA Ames displays (shown at left). They certainly are a standard by which all other cutting-edge systems are judged. These calculate a resolution cone for lateral maneuver. Currently, there is even a 3D veridical version.

Nonetheless, while maps are the standard approach, research data indicate they have trouble optimally representing the full range of maneuver solutions across heading, speed, and altitude. You can get a good lateral OR a vertical maneuver, but it is still extremely hard to get a combination maneuver in manual selection mode.

 

 

My own first attempt at a map display also centered around solving the conflict detection problem. Given the cylindrical geometry of the en-route protected zone ( 5 nm radius, 1000 ft vertical half-height), I wrote a conflict probe to determine if and when one aircraft would enter another's protected zone. This time-to-separation loss was used to color-code the aircraft icons on a map display. The screenshot at left is a 40%-scale view of the actual display we tested with commercial airline pilots. It gave the pilot a way to (1) identify problem aircraft and (2) know how much time there was to resolve the conflict.

While this did solve the conflict detection problem, our research discovered that conflict resolution was the tougher problem to solve. Even when you know which aircraft are threats, you still have to select a maneuver that will solve the conflict. That is the hard part, especially when the intended resolution is either in the lateral (x-y) plane, or involves multiple aircraft.

Several individuals and teams have proposed elegant mathematical algorithms for automatic conflict resolution (the super-autopilot). However, we still feel that strong psychological and political considerations point to having manual conflict-resolution capability--now, and from now on. Pilots want to fly, controllers want to control, and passengers feel comfortable having a pilot onboard. The concept of hurtling through the air in a completely computer-controlled vehicle--while appealing to engineers--is one we feel the public will only accept as a backup system to human guidance.

So, while we know other developers are going to argue vehemently against our position, the issue will be settled in the marketplace. Meanwhile, the job of MST is to explore what we feel is viable technology, and to accurately report on the results.

The problem is the representation

In our opinion, the basic problem with veridical displays is the actual information representation they are based upon, namely physical, geographic reality. As it turns out, while this is the best way to describe the static world, it is not necessarily the best way to describe the dynamic world. This is particularly true when trying to calculate trajectory solutions for multiple objects moving at different speeds, altitudes, and headings.

The need for a new kind of information representation is what motivates the theory of the maneuver space.

To the rest of the world (especially the military), "maneuver space" is defined as the amount of physical space, within which an object can maneuver, given a certain amount of time.

However, it is also possible to define maneuver space as being a conceptual space, namely a space of possible aircraft maneuvers, given a predetermined time period.

The key to this new representation lies in the way we define the word "maneuver." In the simplest sense, any given aircraft maneuver can be defined as one unique setting of the aircraft autopilot. This unique setting consists of 3 numbers, one for heading, one for altitude, one for speed. This defines stable flight (disregarding change from one setting to another, i.e. the actual time it takes to execute a given maneuver--we will discuss that issue presently).

Defined this way, it is possible to construct a 3D volume--a conceptual state space--determined by 3 coordinate axes of heading, altitude, and speed. Individual aircraft maneuvers can then be defined as unique points within that space.

 

Next, it is possible to parse this space into two basic regions--safe maneuvers versus unsafe maneuvers.

Standard mathematical algorithms (the conflict probe--see Footnote) can determine which H,S,A triplets are problematic, as well as the time when separation loss will both occur and end. Any problematic maneuvers can then be color-coded according to time-to-conflict (available maneuver time).

This binary state space now becomes equivalent to answering a set of logical questions: "Suppose I were flying straight and level at heading H, speed S, and altitude A. Would I encounter any conflict with any type of object?" Theoretically, an "object," can be any static or dynamic entity, including traffic, terrain, weather, and/or special-use airspace.

Finally, a pilot can graphically rotate this 3D maneuver space to understand its geometry, and then quickly select a safe, efficient maneuver by simply moving a 3D cursor to a conflict-free region that avoids all conflict regions.

Naturally, this logic has two implicit assumptions (1) a limited conflict probe lookahead time, and (2) instantaneous change from the current H0, S0, A0 to any new point H1, S1, A1 in the state space. Assumption 1 is no special problem because no one can predict reliably into the distant future. All conflict probes assume limited lookahead time anyway.

Assumption 2 is non-trivial, but surmountable, particularly if changes in H,S,A are kept limited in size. First, it is logical to assume that pilots and air traffic controllers prefer efficient maneuvers, defined as those requiring the smallest deviation from current H,S,A. The smaller the deviation, the quicker the maneuver can be executed. Hence, any error due to physical execution time is reduced to manageable limits. Second, future versions of the algorithm will use a more sophisticated conflict probe, eliminating this problem altogether. But, for now, a relatively simple probe should be sufficient to scientifically establish proof-of-concept.