AUTONOMY AS AN ENABLER OF ECONOMICALLY-VIABLE, BEYOND-LINE-OF-SIGHT, LOW-ALTITUDE UAS APPLICATIONS WITH ACCEPTABLE RISK

Proc. AUVSI North America, May 2014

1

AUTONOMY AS AN ENABLER OF ECONOMICALLY-VIABLE, BEYOND-LINE-OF-SIGHT, LOW-ALTITUDE UAS APPLICATIONS

WITH ACCEPTABLE RISK

Ella M. Atkins*

This paper describes a practical vision for low-altitude UAS operations based on a Class

G airspace subdivision that will support safe near-term UAS deployment without impact

to existing manned aircraft operations. An agriculture reference mission is defined as a

case study for which low-altitude UAS offer the landowner tangible benefit without in-

troducing unacceptable risk. A Class U airspace designation is proposed for surface to

500 feet above ground level below existing Class G airspace. Reasonable operational re-

quirements for Class U are shown to significantly depend on overflown property owner-

ship and type. A candidate sub classification of Class U airspace based on property own-

ership (private or public) and type (rural, suburban, and urban) is proposed along with

candidate requirements for the vehicle, its safety features, and its operator(s). Autono-

mous geofencing is proposed as a means to ensure low-altitude UAS do not exit their

designated Class U operating area. A certified geofencing capability can ensure safe

flight in rural areas without the need to wait for system-wide detect-and-avoid so long as

manned aircraft remain clear of the occupied Class U region. A deterministic and simple

geofencing (or electronic leashing) algorithm is presented. Geofencing represents auton-

omy in that it guarantees operating boundaries are respected even if operator inputs must

be overridden. Such algorithms are available today and can ultimately be safety-certified

to provide the backbone autonomy necessary to ensure UAS will not leave their designat-

ed operating area despite flight operations beyond line of sight. The paper concludes

with a discussion of additional technologies needed to achieve adequate safety and priva-

cy management for suburban and urban operations.

INTRODUCTION

Unmanned aircraft system (UAS) missions such as crop inspection will overfly one landown-

er’s property at low altitude. If a UAS can be guaranteed to remain in a box within immediate

reach of the landowner’s property, this UAS does not pose a hazard to aircraft or property outside

this box regardless of whether the UAS is within line of sight (LOS) or beyond line of sight

(BLOS). The expanding “hobbyist” community has embraced this viewpoint by offering autopi-

lots with an “electronic leashing” or “geofencing” option, controversially advertising such autopi-

lots to support beyond-line-of-sight (BLOS) flight capability. Although this technology is not yet

certified for civil use, the idea behind the geofence is compelling and itself suggests a regulatory

strategy by which low-altitude UAS might be supported: create a new classification for very low-

altitude airspace that supports UAS flight in areas currently either considered non-navigable, with

immediate reach of privately-held property, or both.

* Associate Professor, Aerospace Engineering Dept., University of Michigan, 1320 Beal Ave., Ann Arbor, MI 48109.

Proc. AUVSI North America, May 2014

2

The FAA’s recent roadmap for integration of civil UAS1 provides little provision for cost-

effective large-acreage (BLOS) UAS coverage even at low altitudes sufficiently far from airports

to pose no risk to existing manned aircraft traffic. The roadmap instead states that UAS either

must be certified for integrated operation in the NAS, implying high cost and significant delay

associated with certifying detect-and-avoid (previously sense-and-avoid2) and requiring it on all

aircraft, or that operations must occur within line-of-sight (LOS), not a practical constraint for

large-area BLOS applications. Behind this position is the ICAO-backed hypothesis that all UAS

are remotely-piloted aircraft (RPAs) without accommodation of fully-autonomous operations.

This paper investigates how a modest rulemaking effort focused on reclassifying low-altitude

uncontrolled airspace and capitalizing on a simple, deterministic3 autonomy algorithm for

geofencing can enable safe, cost-effective BLOS UAS operations. A new Class U low-altitude

airspace is proposed and assigned common-sense operational rules based on overflown property

ownership (private vs. public) and type (rural vs. suburban vs. urban). An agriculture case study

is proposed as an initial means of enabling commercial UAS in a manner that poses risk to neither

people nor property. Subsequent progression toward more populated and publically-owned land

is then described. The first message of this paper is that low-altitude airspace rules must take the

characteristics of overflown property into account – otherwise operational approval will either be

unsafe (e.g., allowing unlimited hobbyist flight operations in urban areas) or repressive (e.g., pro-

hibiting agricultural flights at low-altitude over privately-owned rural property). The second

message of this paper is that autonomy without possibility of pilot override can actually enable

safer flight, particularly for BLOS operations where human perception is achieved only through

graphical ground station interfaces. Geofencing is a specific example of an autonomy technology

specifically focused on keeping a UAS within its designated operating area. With a trusted

geofence, loss of communication link will not result in emergency flight termination (ditching)

and will not allow exit from the test range. Waypoint entry errors also will not cause geofence

violation unless the operator enters both waypoint and the geofence data incorrectly.

The paper is organized as follows. Below, UAS agricultural operations are introduced as a

motivating case study, including a discussion of platform types, operational paradigms, and legal

precedent that collectively must be considered in UAS regulatory policy. Next, a new class of

low-altitude airspace (Class U) is proposed and discussed in the context of property attributes and

operating requirements. Geofencing to ensure airspace constraints are respected is then dis-

cussed, focusing on a simple deterministic algorithm available today in an open-source autopilot

to provide context on how autonomy might be realized without introducing nondeterminism or

unmanageable system complexity. The paper concludes with a summary and discussion of future

work required to move forward in the suite of technological and policy challenges required to

comprehensively support UAS operations.

UAS FOR AGRICULTURE: A RECOGNIZED AND MOTIVATING CASE STUDY

There are numerous low-altitude UAS applications, some necessitating flight over densely

populated areas. Perhaps the “killer app” for the near term, however, is agriculture. First, low-

altitude UAS flights over privately-owned rural property will not pose a risk to people or property

on the ground other than those who own the land (and have chosen to fly). Unless the property is

in close proximity to an airport, airspace at low-altitude is currently uncontrolled (Class G) and

rarely to never occupied by manned aircraft. These circumstances imply very low-altitude UAS

operations over rural farms are safe whether the vehicle lands intact or not so long as the vehicle

does not depart its low-altitude operating range or “box” over the privately-owned farm. Low-

altitude UAS flights over privately-owned farmland also cannot raise [valid] privacy concerns so

Proc. AUVSI North America, May 2014

3

long as these flights are conducted by or approved by the landowner, who would then have exclu-

sive rights to any collected data including the right to decide how this data is disseminated.

Agricultural UAS flights would serve two primary purposes: surveillance and chemical appli-

cation, as shown in Figure 1. Crop, livestock, or infrastructure (e.g., fencing) inspection by air

can significantly improve situational awareness for the farmer at reasonable cost. Such awareness

will enable irrigation and pesticide application to be performed more effectively and efficiently,

find or survey livestock and their food/water sources, and identify specific problems with fencing.

Pesticide application by UAS can reduce risk to crop duster pilots due to pesticide exposure as

well as the relatively dangerous flight pattern flown to minimize operational cost. The Yamaha

RMAX is under consideration for agriculture operations in the USA

(http://www.ainonline.com/aviation-news/2013-09-23/after-first-commercial-flight-other-uas-

types-advance), having already established a track record for successful crop spraying overseas

(http://rmax.yamaha-motor.com.au/history).

(a) CropCam for imaging (b) Yamaha RMAX for pesticide application

(http://www.robotshop.com/en/cropcam-unmanned-aerial-vehicle-uav.html) (http://www.suasnews.com/2014/01/26971/unlocking-the-potential-of-unmanned-aircraft-systems/)

Figure 1. Commercial UAS for agriculture applications.

At first consideration, one might think establishment of a ruling enabling small UAS flight

within line-of-sight (LOS) will be sufficient to support agriculture; however, this is not the case.

First, note that pesticide spraying craft such as the RMAX are necessarily heavier than what we

typically consider “small” UAS so it is important that risk be appropriately assessed4 which in the

case of rural agriculture does not require establishing specific size or weight constraints. Second,

the agriculture surveillance UAS will need greater flexibility than a LOS SUAS rule would offer.

Consider the powered glider and imaging UAS such as that shown in Figure 1a. This platform

will have a range and endurance that can and should support beyond-line-of-sight (BLOS) flight.

BLOS flight is essential for such craft to be effective, as many farms capable of appreciable profit

have large acreage that extends well beyond human visual range. Further, some crops can be ob-

served more effectively by alternative instrumentation (e.g., thermal, infrared scanners) at night

due to stomata closing to reduce evapotranspiration during the day. Perhaps the true “killer app”

for crop inspection by UAS is to fly the powered glider BLOS at night. Such an operation would

be most cost-effective and user-friendly if the farmer could deploy the craft just before retiring

for the evening. The UAS would then fly its pre-programmed mission overnight, downloading

status and alert data throughout the flight. The UAS would end its flight by autonomously return-

ing to base and (optionally) downloading collected data before the sun rises. The farmer could

then get a good night of sleep and examine collected data the next morning.

Proc. AUVSI North America, May 2014

4

There are numerous benefits to the “overnight crop inspection” mission. First, collected data

may be more informative due to night-time crop state as mentioned above. Second, night-time

flight in a rural area is even less likely than daytime flight to pose risk to people on the ground or

in the air. Third, given that profit margins are already small for most agricultural operations, au-

tonomous flight frees the farmer or any employees to complete the usual farm work without the

need to somehow find several “free hours” to monitor a UAS ground station display.

As will be discussed below, for low-altitude UAS applications to be realized reasonable guide-

lines must be established with respect to both airspace and autonomy. As a baseline for the rural

agriculture application, farmers simply need recognition of their freedom to fly in a manner that

can beneficially impact their livelihood. This can be accomplished by rulemaking efforts that

will take significant time at potential loss of US competitiveness, or else through recognition of

the private property owner’s right to occupy the airspace within immediate reach of their property

without regulatory oversight.

A precedent for “immediate reaches airspace ownership” by the property owner has in fact al-

ready been established by the Supreme Court (United States v. Causby – 328 U.S. 256 (1946),

http://supreme.justia.com/cases/federal/us/328/256/case.html). In this case, a landowner and

farmer argued that low-flying aircraft were causing harm to his chickens thus his livelihood. The

Supreme Court ruled that the landowner did have ownership to some volume of “immediate

reaches” airspace above his land, but the Court did not offer a specific above-ground-level alti-

tude defining the extent of immediate reaches airspace. Excerpts from the ruling include: “We

have said that the airspace is a public highway. Yet it is obvious that, if the landowner is to have

full enjoyment of the land, he must have exclusive control of the immediate reaches of the envel-

oping atmosphere. Otherwise buildings could not be erected, trees could not be planted, and even

fences could not be run.” (Page 328 U.S. 264), and “The superadjacent airspace at this low alti-

tude is so close to the land that continuous invasions of it [by publically-regulated aircraft] affect

the use of the surface of the land itself. We think that the landowner, as an incident to his owner-

ship, has a claim to it, and that invasions of it [by publically-regulated aircraft] are in the same

category as invasions of the surface.” (Page 328 U.S. 265). Quite simply, UAS use within imme-

diate reaches of privately-owned (rural) property can impact the owner of said property for agri-

culture and other uses. Such impact must therefore be carefully considered in the context of the

United States v. Causby ruling. If the landowner wants UAS flights he/she should be free to con-

duct them or grant permission for others (e.g., commercial operators) to conduct such operations.

Conversely, if the landowner does not want any UAS operations over his/her property, he/she

should also be free to say “no” to UAS presence in his/her immediate-reaches airspace. Recogni-

tion of immediate reaches airspace will support the emerging commercial UAS industry while

simultaneously addressing privacy concerns. This is also consistent with United States v. Causby.

The agriculture applications described in this section may perhaps be accomplished without

specific regulatory oversight except formal definition and recognition of low-altitude airspace

over privately-owned property as “immediate reaches” (thus not taken for public use), “non-

navigable” (in that it is sufficiently low or close to obstacles to pose a hazard at least to manned

fixed-wing aircraft),4 or both. Care does need to be taken that such operations will remain within

designated lateral and altitude airspace boundaries, much like farmers are responsible for keeping

livestock within fenced property boundaries (but perhaps with a higher level of confidence for the

UAS). BLOS UAS operations, day or night, will necessarily rely on instrumentation and soft-

ware to remain within designated boundaries. Geofencing systems, already available on some

UAS, can provide such a capability through autonomy that can and will override a pilot who di-

rects the aircraft outside property bounds. Specifics on potential airspace designations and the

necessary autonomy required for robust and certifiable geofencing are discussed below.

Proc. AUVSI North America, May 2014

5

CLASS U: PRACTICAL LOW-ALTITUDE AIRSPACE ALLOCATION FOR UAS

The National Airspace System (NAS) is currently divided into a set of classes (A, B, C, D, E,

G) that distinguish operational requirements. The NAS has a “static” structure that is specified on

aviation charts as shown in Figure 2 with the operational requirements summarized in Figure 3.

As shown, very low-altitude airspace is designated as uncontrolled (Class G) except near pub-

lished airports. Commercial transport aircraft and their passengers primarily occupy airspace

Class A enroute and Class B for approach and departure. Commercial and charter flights also

operate out of smaller airports surrounded by Class C or Class D airspace. General aviation

flights typically operate out of smaller airports (Classes C, D, E) and then typically fly above the

Class E airspace floor to minimize risk due to terrain proximity. Exceptions to this rule include

missions involving overflight of a particular property (e.g., aerial inspection or pesticide applica-

tion). For fixed-wing aircraft, altitudes less than 500’ above ground level (AGL) are not consid-

ered safely navigable per the Federal Aviation Regulations (FARs)* thus should never be occu-

pied except on final approach and initial departure where obstacles must be mapped and marked.

Figure 2. The National Airspace System (NAS): Current classification. (https://www.aopa.org/-/media/Files/AOPA/Home/Pilot%20Resources/ASI/various%20safety%20pdfs/airspace2011.pdf)

Rotorcraft (helicopters) may operate within or transit Class G airspace. However, rotorcraft

have no real need for extremely low flight (e.g., less than 500’ AGL) over rural private property

unless performing a law enforcement operation with appropriate search warrant or emergency

service mission5 (e.g., search-and-rescue, medical evacuation, firefighting) for which landowners

would readily ground their own UAS and grant temporary and immediate permission to overfly

and also to land as needed. In urban and suburban areas rotorcraft typically fly across a series of

public and private properties necessitating rulemaking to safely support existing manned and new

unmanned aircraft operations. Once approved it is likely that commercial low-altitude UAS oper-

ations will soon outnumber low-altitude manned operations of any type;† the question is how to

support both safely and effectively. Clearly the answer cannot be to continue labeling all low-

altitude airspace as Class G (uncontrolled) as this creates the current chaotic environment in

* Per 14 CFR 1.1 General Definitions: “Navigable airspace means airspace at and above the minimum flight altitudes

prescribed by or under this chapter [14 CFR], including airspace needed for safe takeoff and landing.” Per 14 CFR

91.119 Minimum safe altitudes: provided no emergency landing is required, aircraft must operate above 500’ AGL

(rural) or 1000’ AGL (congested areas including urban, suburban, and other assemblies of persons). Exceptions are

given to helicopters, powered parachutes and weight-shift-control aircraft. The latter two types tend to operate locally

thus would not overfly many different properties unless [unwisely] operating over small-tract urban/suburban areas. † In fact, the number of LOS hobbyist operations per day already likely outnumbers the typical number of manned ro-

torcraft flight operations per day.

Proc. AUVSI North America, May 2014

6

which manned rotorcraft attempt to see-and-avoid, hobbyists fly anywhere they please, and com-

mercial operators are told not to fly.

Figure 3. Operational requirements by airspace class. (https://www.aopa.org/-/media/Files/AOPA/Home/Pilot%20Resources/ASI/various%20safety%20pdfs/airspace2011.pdf)

This paper proposes introduction of a new airspace class, say Class U, that establishes the sort

of common-sense requirements to enable low-altitude UAS that mirror the existing airspace clas-

ses that historically have made a great deal of sense for manned operations. As a start, the new

Class U designation could be exclusively defined as the airspace from ground level up to, say,

500’ AGL, directly below all existing Class G airspace. With this definition the new Class U

airspace would have absolutely no impact on commercial transport operations* and almost no im-

pact on other manned operations with the exception of the rotorcraft missions discussed above.

Figure 4 illustrates the proposed Class U airspace designation, shown in yellow under all Class G

airspace. While this appears as one airspace type, appropriate policy will require further sub clas-

sification by attributes such as property ownership and type. This additional complexity in classi-

fication is required because operations in Class U will necessarily occur close to the ground

where operational approval criteria, level of risk, and potential for privacy violation are strongly

influenced by the attributes of the property being overflown.

ClassB

Class E

Class U(*,*)

Figure 4. Class U airspace (<500’ AGL).

The (*,*) notation indicates property ownership and type as shown in Table 1.

* Provided Class U occupants did not illegally enter other airspace.

Proc. AUVSI North America, May 2014

7

Table 1 proposes a specific sub classification for Class U airspace. Property can either be pri-

vately or publically owned, while property type can be rural, suburban, or urban. Collectively

these attributes generate six distinct sub classifications. Although a quantity of six sub classes

may seem a bit cumbersome from a regulatory perspective, the diversity in operation types makes

it absolutely essential for rulemaking to reflect the appropriate needs and concerns for each class

of property. Without such distinctions, common-sense rulemaking will be impossible. To sup-

port this claim, consider two of the example missions listed in Table 1: flight over a national for-

est versus flight through an urban canyon. For these examples, both properties are publically-

owned thus subject to rulemaking rather than immediate-reaches/imminent domain per the above

agriculture case study. Low-altitude overflight of a national forest is challenging because of po-

tentially complex terrain and obstacles (trees), but loss-of privacy and safety risks to people and

[improved] property on the ground are minimal so long as the UAS doesn’t create a forest fire,

stalk tourists, or spill chemicals following an emergency flight termination. Conversely, flight

down an urban canyon supports the well-publicized “delivery drone” mission with high potential

to generate capital, but this mission is fraught with situations that place both people and property

at risk. Both payload and flight sensors also have potential to violate privacy in the urban can-

yon. Rulemaking must directly deal with these differences if common-sense policy is to emerge.

Table 1. Class U Airspace Overflown Property Designations.

Airspace

Class

Property

Class

Property

Ownership

Property Type

(Population)

Examples

U (P,R) Private (P) Rural (R) Farm, utility (e.g., pipeline), unimproved (e.g.,

forest, desert, mountain, water)

U (P,S) Private (P) Suburban (S) Residential, business

U (P,U) Private (P) Urban (U) Multi-tenant residence, business

U (G,R) Govt / Public (G) Rural (R) National park, national forest, waterway

U (G,S) Govt / Public (G) Suburban (S) Transportation infrastructure, public utility

easements

U (G,U) Govt / Public (G) Urban (U) Public parks, public buildings, city streets,

bridges

Table 2 proposes an initial draft of operational requirements for the six Class U property-

based sub classifications. This draft would of course require substantial analysis and revision to

ensure an adopted policy is appropriate. The fields in Table 2 intentionally mirror those in Figure

3 to facilitate comparison. New link requirements and safety requirements achieved primarily

through onboard automation technology are also proposed. As shown, the privately-owned, rural

operation has no link or pilot training requirement, consistent with the concept that the landowner

or lessee with explicit landowner permission is responsible for authorization and safety of UAS

operating in his/her “immediate reaches” airspace. For Class (P,R) the only safety requirement is

a geofence capability to ensure the UAS does not exit the private property owner’s airspace.

While both of the remaining privately-owned property classes require owner permission or a

search warrant for entry, additional requirements are prudent because these operations will occur

on smaller-acreage properties, e.g., residential neighborhoods. Confined operations over private-

ly-owned properties, e.g., for real estate videos, can likely be accommodated through local link.

Safety features such as the geofence plus flight termination are prudent given the small test range,

and line-of-sight (LOS) restrictions consistent with the local link are reasonable. This type of

operation is already being conducted by “hobbyists” with only basic transmitter “stick skills”. It

is reasonable to expect that such LOS operations can continue to be conducted safely with train-

Proc. AUVSI North America, May 2014

8

ing consistent with that recommended by AMA (the Academy of Model Aeronautics); it is less

reasonable to require an LOS operator of Class U (P,S) aircraft who will likely operate the UAS

while standing in their back yard or sitting in their lawn chair to pass an aviation medical exam

and/or earn a manned aircraft pilot’s license. For urban operations over privately-owned proper-

ty, the expectation is that this property size will be smaller (e.g., above a high-rise roof), so an

additional VTOL (vertical takeoff and landing) restriction may be prudent.

Flight above publically-owned property will necessarily be governed through rulemaking,

perhaps a combination of federal, state, and locally-specified guidelines. Examples distinguish-

ing “common-sense” differences between types of public property overflown were given in Table

1, which in turn suggest the requirements listed in Table 2 below. Low-altitude flight over rural

areas, whether private or publically-owned, does not pose risk to people or property. However,

any operator should expect to assume responsibility to recover his/her UAS if it does not return to

the designated recovery site. To support this need, Table 2 recommends return-to-base and flight

termination capabilities as well as “facility training” for the UAS operator. Note that facility

training would be aimed at ensuring the operator understood how to operate safely above and on

the public property in additional to having AMA-level skills. Suburban and urban flight over

publically-owned properties, including roads, bridges, waterways, and parks, will require the

most regulatory oversight of the proposed six Class U sub classes. Both (G,S) and (G,U) opera-

tions have potential to pose loss-of privacy and safety risks to people and property. Additional

requirements for (reliable) long-range communication, UAS pilot certification, and detect-and-

avoid (with ground and air) are proposed. Note that the designation “UAS pilot” implies training

that leads to certification. A competent UAS pilot will require understanding of their craft, in-

cluding link and automation behaviors. The UAS pilot will also need to know how to interact

with ATC and manned aircraft (e.g., rotorcraft) that might occupy shared airspace. However,

specifics on the UAS pilot rating require further research; such a rating must not be presumed to

require either an aviation medical exam or any manned aircraft pilot license as the UAS pilot will

never be exposed to low-density atmosphere or high-g maneuvers and will not have the immer-

sive sensory feedback available to an onboard pilot.

Table 2. Candidate Operational Requirements for Each Class U Property Sub Class.

Property

Class

Entry Requirements Link

Requirements

Safety

Requirements

Minimum Pilot

Qualifications

(P,R) Owner permission or

search warrant

None Geofence None

(P,S) Owner permission or

search warrant

Local Class U(P,R), flight termina-

tion, LOS

AMA

(P,U) Owner permission or

search warrant

Local Class U(P,S), VTOL AMA

(G,R) Public agency permis-

sion

None Class U(P,R), flight termina-

tion, return-to-base

AMA, facility train-

ing

(G,S) Community permis-

sion, certification

Long-range Class U(G,R), detect-and-

avoid (ground)

UAS pilot

(G,U) Municipality permis-

sion, certification

Long-range Class U(G,S), VTOL, detect-

and-avoid (air)

UAS pilot

While most low-altitude missions can be conducted in the proposed permanent Class U air-

space, some missions may require flight above the Class U boundary. Certainly, comprehensive

integration including detect-and-avoid will enable such flights but this is a very long-term solu-

Proc. AUVSI North America, May 2014

9

tion as most general aviation and low-cost UAS will not be equipped with certified detect-and-

avoid for some time. An alternative to full detect-and-avoid is dynamic airspace allocation. In

the modern age of connectivity, Class U airspace “boxes” can be reserved and carved from Class

E and G airspace. Such usage can be communicated system-wide through the web / cell network

or means such as the NOTAM system plus regular announcements on local ATC voice communi-

cation frequencies.

ClassB

Class EClass U(t0,tf)

PermanentClass U

under Class G

Dynamic allocation of (temporary) Class U from Class E or G

Figure 5. Dynamic Class U airspace allocation.

ELECTRONIC GEOFENCING (OR LEASHING) FOR SAFETY MANAGEMENT

The above discussion proposes a Class U airspace and suggests its progressive introduction,

starting with the “easiest to implement” rural, privately-owned sub class. As shown in Table 2,

flight in privately-controlled (immediate reaches) airspace will be safe to entities outside this pri-

vately-controlled airspace so long as each aircraft (UAS) remains within designated altitude and

lateral airspace boundaries. Ensuring a UAS remains within designated boundaries, especially

when flying BLOS and/or night-time operations, requires a geofencing or electronic leashing au-

tomation aid. This automation aid in reality represents autonomy in that the pilot-in-command

(UAS operator) will not have full override authority should said pilot be directing the UAS to exit

the designated Class U airspace when the geofencing software intervenes to ensure the aircraft

remains within the designated area.*

Figure 6 shows a candidate geofencing algorithm that is straightforward in the software and

for the operator. This algorithm is a high-level summary of the geofence software implemented

in the open-source Ardupilot / Arduplane system recently repackaged as the APM Autopilot suite

with upgraded processing capabilities. Software-based geofencing is optional in this system – the

operator can choose no geofence, altitude-only, or lateral plus altitude. Because the Ardupilot is

intentionally simple, the geofence never prompts replanning of the flight – it instead triggers de-

scent and/or a turn back to a central “base” or “return-to” waypoint.

* Common arguments against autonomy involve “what if…” (anomalous) scenarios. UAS operating over private prop-

erty can flight terminate at any juncture with risk only to the private property and its occupants; a property owner’s

decision to fly (or do anything else for that matter) on or immediately above their own property is always at their risk.

Proc. AUVSI North America, May 2014

10

Function geofence (altitude_check_only_flag, geofence_violation_flag)

1. Check if geofence is enabled and load any new geofence definition. a. If geofence is not enabled, check if pilot/operator has activated

the geofence control reset switch, reset switch per direction from

operator, and RETURN.

b. If geofence is enabled, check for new boundary definition and load as needed. CONTINUE if new geofence is enabled; otherwise RETURN.

2. If geofence minimum or maximum altitude is violated (new violations only; do not trigger on existing violations indicated by a set violation_flag):

a. Set new reference/waypoint altitude to return value (either user-specified if acceptable or exactly between minimum and maximum

values otherwise).

b. Activate autopilot altitude flight control law as needed.

3. If altitude_check_only flag is false and geofence lateral boundary lati-tude, longitude is violated (new violations only; do not trigger on ex-

isting violations indicated by a set violation_flag):

a. Set lateral latitude, longitude waypoint to a default “return-to-base” value.

b. Activate autopilot lateral flight control law as needed.

4. If operator has triggered a reset, reset violation_flag to zero and re-sume regular manual or waypoint flight.

Figure 6: Ardupilot Geofencing Summary (http://ardupilot.com/).

Processor

DatalinkAutopilotGeofence

Processors

DatalinkAutopilot

Pilot Controls

Laptop

Ground station

UAV

Geofence

UAV

Pilot Controls

Laptop

Ground station

Servos SensorsServos Sensors

(a) Single processor architecture (b) Multi-processor architecture

Figure 7: Candidate geofencing processor architectures.

Figure 7 shows the two basic processor configuration strategies for implementing a software

geofencing capability. On the left, a single onboard processor handles all nominal functionality

plus geofencing. This is desirable and was the default Ardupilot (with Arduino) configuration

because a single processor minimizes cost, weight, power, and wiring harness complexity. On

the other hand, if other software on this shared processor experiences a problem, the geofence

Proc. AUVSI North America, May 2014

11

also will fail, suggesting an architecture (Figure 7b) where the geofence software is pulled onto a

separate (micro)processor with the sole purpose of ensuring the vehicle remains within its desig-

nated operating area. The Figure 7b design is likely best for research groups who will be modify-

ing nominal autopilot functionality thus potentially introducing software problems; the Figure 7b

design can also provide computational redundancy so long as both processors nominally guide

the aircraft inside its operating area, although for full redundancy the processors would need to

utilize independent servo connections and inertial navigation sensors as well as providing logic

such as a watchdog-triggered bypass for servo output source selection. A key feature of the Fig-

ure 7b design is the possibility of certifying only the geofencing module independent of the rest

of the autopilot system, which should be sufficient for safety certification in applications such as

LOS test flights and rural agriculture.

CONCLUSION AND FUTURE WORK

This paper has described a practical vision for low-altitude UAS operations that will enable

their immediate (or near-term) deployment through judicious use of autonomy to assure safety

using techniques such as software-based geofencing. Agriculture is defined as a compelling ref-

erence mission for low-altitude UAS that must operate above the “small” category and BLOS to

be effective. A Class U airspace designation is proposed for surface to 500’ AGL below airspace

currently designated as uncontrolled (Class G). Operational requirements for Class U necessarily

and for the first time depend on property ownership and type to ensure safety and privacy while

also supporting rather than unnecessarily restricting commercial UAS operations. Autonomous

geofencing is proposed as a means to ensure low-altitude UAS do not exit their designated Class

U operating area. A certified geofencing capability can in fact ensure safe flight in rural areas

without the need to wait for system-wide detect-and-avoid capability so long as manned aircraft

remain clear of Class U or else obtain specific authorization thus clearance through specific pri-

vate or publically-owned Class U airspace regions.

For urban areas, the geofence must be supplemented by additional safety and even privacy

management capabilities. Perhaps most important is detect-and-avoid not just for other aircraft

but also for people and property on the ground. Given the need to collect a large volume of sen-

sor data both to ensure safe operations and to support surveillance missions, policy to ensure ap-

propriate privacy of collected data is indeed needed for urban and potentially suburban opera-

tions. Privacy must be specifically recognized as not being a concern for rural UAS operations

conducted by private property owners; otherwise privacy issues will continue needlessly delaying

approval of these rural operations.

While Class U airspace could be created in the near-term through a concerted rulemaking ef-

fort, it will be prudent to phase in deployment of Class U from the “easiest” case of private prop-

erty, rural (P,R) to the more challenging case (G,U) for which the public expects privacy and de-

mands safety. Additionally, public entities (e.g., municipalities, department of transportation au-

thorities) must commission or at least approve UAS overflight of their properties. Such entities

must gain a better and more realistic understanding of the needs and desires of their constituents

with respect to benefits of the operations as well as safety and privacy concerns. Only then can

these needs and desires be translated to policy that makes sense. Urban UAS operators will then

Proc. AUVSI North America, May 2014

12

only be allowed to “darken the skies” with delivery drones should their activities truly be safe and

supported overall by the community being overflown.*

This paper motivates and proposes a regulatory scheme to enable low-altitude UAS operations

with neither the need for manned aircraft detect-and-avoid nor real-time air traffic control over-

sight. This strategy will enable commercial UAS operations and will finally enable both com-

mercial and academic entities to routinely flight test emerging technological capabilities. How-

ever, this operational paradigm is incomplete. Substantial additional research, procedure, and

policy developments are required to accomplish the end-goal: comprehensive integration of UAS

into the NAS. It is this author’s conviction, however, that the proposed Class U airspace will

continue to fill a long-term rather than a short-term need, as even when technology and training

are ready to accommodate full integration, property ownership and type must still factor promi-

nently into low-altitude airspace ownership, operational approval, certification, and training re-

quirements.

ACKNOWLEDGMENTS

The author would like to thank Jerry Lin for his work to decode and flow chart the Ardupilot

geofencing algorithm. This work was supported in part by NASA contract NNX11AO78A.

REFERENCES

1 Federal Aviation Administration (FAA), Integration of Civil Unmanned Aircraft Systems

(UAS) in the National Airspace System (NAS) Roadmap, US Department of Transportation,

First Edition, 2013. http://www.faa.gov/about/initiatives/uas/media/uas_roadmap_2013.pdf

2 X. Prats, L. Delgado, J. Ramírez, P. Royo, and E. Pastor. "Requirements, Issues, and Chal-

lenges for Sense and Avoid in Unmanned Aircraft Systems," Journal of Aircraft, AIAA,

Vol. 49, No. 3 (2012), pp. 677-687.

3 J. Neidhoefer, C. Gibson, M. Kunda, and E. Johnson. "Determinism and Autonomy in the

National Airspace System (NAS)," Journal of Aerospace Computing, Information, and

Communication, AIAA, Vol. 4, No. 11, (2007), pp. 1037-1045.

4 A. Lacher and D. Maroney, “A New Paradigm for Small UAS,” Proc. AUVSI North America,

Las Vegas, NV, August 2012.

5 K. Gatian, H. Jimenez, and D. Mavris. “Requirements Analysis of Unmanned Aircraft Sys-

tems for Emergency Response Operations,” Aviation Technology, Integration, and Opera-

tions (ATIO) Conference, AIAA, Indianapolis, Indiana, Sept. 2012.

* Such endorsement will likely involve financial compensation in the form of operator taxation and/or public airspace

usage fees, not just the public’s overwhelming desire to have packages or pizzas delivered by air.