Content uploaded by Siddharth Agarwal
Author content
All content in this area was uploaded by Siddharth Agarwal on Jul 19, 2019
Content may be subject to copyright.
Characteristics of Indoor Disaster Environments for
Small UASs
Siddharth Agarwal
Electrical and Computer
Engineering
Texas A&M University
College Station, TX 77843
Email: siddharth@tamu.edu
Robin R. Murphy
Center for Robot-Assisted
Search and Rescue
Texas A&M University
College Station, TX 77843
Email: murphy@cse.tamu.edu
Julie A. Adams
Electrical Engineering and
Computer Science
Vanderbilt University
Nashville, TN 37067
Email: julie.a.adams@vanderbilt.edu
Abstract—This paper provides a formal analysis of indoor
disaster environments that impact the design of small unmanned
aerial systems (SUASs), their navigational algorithms, and their
sensors. Four characteristics of a region of space: scale, degree
of deconstruction, location of obstacles, and tortuosity are de-
scribed. The analysis compares the value of these characteristics
for Prop 133 at Disaster City R
with twelve SUASs that have
flown inside a disaster damaged building or other physical or
computer simulated indoor space; the analysis normalizes the
platform size. Eleven of the twelve systems were tested in more
spacious regions (habitable) than the regions typified by Prop
133 (restricted maneuverability). Only one of the twelve systems
was tested in a deconstructed environment; likewise only one
testbed placed obstacles in equivalent configurations to those in
Prop 133. The tortuosity of the testbeds was at best half of the
tortuosity of Prop 133. The paper concludes that current obstacle
avoidance and simultaneous localization and mapping algorithms
vetted with those testbeds may not perform well in actual disaster
environments.
I. INTRODUCTION
The possibility of using small multi-rotor unmanned aerial
systems (SUASs) for surveying damage inside buildings and
structures is increasing. The successful flights by a University
of Pennsylvania/Tohoku University team [1] in a multi-story
building damaged by the 2011 Great Eastern Japan Earthquake
and the multi-university NIFTi team’s inspection of cathedrals
in Mirandola collapsed by the Finale Emilia Earthquake [2]
demonstrate the potential utility of SUAS for multi-story
buildings and processing facilities, such as Fukushima Daiichi.
However, the inability of SUAS to make progress in indoor
flights inspecting buildings in Biloxi damaged by Hurricane
Katrina in 2005 [3] and the Christchurch Catholic Basilica
damaged by the 2011 Christchurch, New Zealand, Earthquake
[4] act as reminders of remaining challenges.
Small UAS for flying indoors is fortunately an active area of
investigation [1], [5], [6], [7], [8], [9], [10], [11], [12], [13],
[14], [15]; however, advances in flying for normal, undam-
aged indoor environments may not be directly transferrable
to disaster response. Kinetic events, such as earthquakes,
tornadoes, hurricanes, industrial accidents, or explosions often
deconstruct interiors, while leaving the building compromised.
A mild earthquake may rearrange office furniture, knock over
bookcases, and cause ceiling fixtures to hang loose, while
leaving the structural elements, such as walls, ceilings, floors,
and pillars intact. A more severe event will have a higher
degree of deconstruction, collapsing walls and ceilings, de-
positing debris, and changing the overall layout of the building.
Indoor SUAS navigational algorithms for obstacle avoidance
and simultaneous localization and mapping (SLAM) depend
upon assumptions about the environment that influence the res-
olution of maps, the choice of sensors and sensor processing,
and the update frequency of control and sensing. Therefore,
having an accurate characterization of deconstructed indoor
environments is essential to developing indoor SUAS that can
fly in realistic disaster conditions.
The approach taken in this paper is to first define a
characterization of indoor deconstructed environments for a
SUAS in terms of scale, degree of deconstruction, severity of
obstacles, and tortuosity [4]. The Prop 133 at Disaster City R
is presented as a pictorial example of these characteristics for
an average case. While Prop 133 stages only one possible
scenario, the partial collapse of a multi-story office building,
it is a realistic representation used for training responders
and thus is helpful in visualizing the definitions. The paper
summarizes twelve studies of indoor SUASs and compares
the environmental characteristics explored by these systems,
versus the average disaster case. The comparison produces
conclusions as to remaining gaps in the state of the art in
indoor SUAS for disasters.
A. Definitions
This paper uses the definitions from Disaster Robotics
[4]. While these definitions were originally developed for
unmanned ground vehicles, they can be extended to the
three dimensional environments SUAS operate within. The
operational envelope for a SUAS is defined as a collection
of one or more regions. For example, in a multi-story office
building, a hallway is a distinct region from an office and a
stairway. The navigability of a specific region for a SUAS can
be described in terms of scale and traversability.
The scale of a region reflects the relationship of the size of
the agent Ato the size of the environment E[4]. An agent can
be a person or robot. A large environment, such as high bay
provides more space for a SUAS than a narrow hallway. To
quantify this, scale is given as the relative size of characteristic
dimension CD of the agent and environment. The Acd is
the largest single dimension affecting SUAS navigation. For
example, a multi-rotor SUAS may have an overall platform
size with a diameter of 0.5m in the horizontal plane with
cameras and payloads protruding 0.2m, and a constraint that
the SUAS is never allowed closer than 0.3m to an obstacle.
Therefore, the maximum dimension is 0.5m+ 0.3m= 0.8m;
thus, Acd = 0.8m. Note that the Acd is the equivalent
of reducing a SUAS to a sphere. The Ecd is the nominal
minimum dimension of the environment affecting navigation.
For a hallway, it is the average width, as obstacles intruding
into the hallway are rare. An office may have a smaller Ecd,
where the furniture is arranged to allow a human to walk
through, but with less free space than in a hallway and a lower
ceiling.
The intrinsic navigability of a region based on scale can
be categorized as one of three indoor regimes in [4]. In
the habitable regime, Ecd >2Acd, the agent can move
freely through the environment. For a human, this regime
represents “normal” interior spaces designed for people to
work and live in that have not been altered by a kinetic
disaster. For a SUAS with a Acd about the width of a
person, a human habitable space will be the same as a SUAS
habitable space. A SUAS may be deployed into a habitable
environment if there was a chemical, radiological, or biological
incident where human movement was restricted by safety
procedures or personal protection gear, such as the use of
UGVs at the Fukushima Daiichi nuclear emergency. In the
restricted maneuverability regime, Ecd <2Acd, the agent
can still move in the environment, but that movement is
restricted by the much narrower spaces. The environment may
be naturally small, such as a sewer pipe; however, the more
interesting case for disasters are human habitable environments
that have become deconstructed from normal dimensions. A
partially collapsed building from a kinetic event, such as an
earthquake or explosion that a responder can walk through,
though perhaps have to bend over or squeeze through, is an
example of a restricted maneuverability regime. Robots for
surface entry into mine disasters or parking garage collapses
function in this regime. In the third indoor regime, the agent
is burrowing into the environment and working at a granular
level, Ecd < Acd. It is not possible for an SUAS to displace
material and create space for itself, so only the habitable and
restricted maneuverability regimes are discussed.
The traversability of a specific region by a SUAS can be
decomposed into four primary environmental characteristics.
The verticality of the region refers to the nominal nature
of flight. A stairwell region has high verticality, while a
hallway may have a high ceiling, but in a building the SUAS
will be primarily flying horizontally through a region, not
vertically. The degree of deconstruction of an indoor region
reflects the condition of the structural elements, essentially
are the walls and ceilings still orthogonal and in place. The
severity of obstacles captures the number and size of obstacles
that temporarily reduce the Ecd and may require obstacle
(a)
Room 1 Room 2
Room 3 with Ceiling Lean-to
Collapse
(b)
Fig. 1: Prop 133 at Disaster City R
. a) View of the federal
building component and b) the floor plan for the first and
second floors.
avoidance. If the environment is essentially a path through
nearly continuous obstacles, the free space between obstacles
becomes Ecd. A normal habitable space will have very few
navigational obstacles, as human spaces are designed for
people to move and work in. A kinetic event may deconstruct
the habitable space by creating debris and hanging obstacles.
The deconstruction and severity of obstacles in turn leads to
the tortuosity of a region. Tortuosity represents the meters
between turns, including changes in altitude, in the region for
navigation; it does not include yawing to provide sensor views.
In addition, the performance of an indoor SUAS will also
be influenced by other secondary components. The lighting
of a region is likely to be variable, which can influence
sensing. The surface properties of the regions refer to the
materials present in the environment. These materials can
impact sensing, for example, acoustic ceiling tiles absorb
ultrasound signals thereby distorting the response.
B. Disaster City R
Since SUASs have been used only four times for surveying
the interior of damaged buildings [4], those data sets are
too limited to project the broad set of regions for indoor
flight. However, Prop 133 at Disaster City R
was designed
to represent an average expected state of a damaged multi-
story commercial building. Disaster City R
is a complex of
props designed by professional trainers, who are themselves
responders, to accurately represent physical conditions that
urban search and rescue (US&Rs) teams will experience for a
range of disasters. It is owned by the Texas A&M Engineering
Extension Service and is used to train over 80,000 humans
and canines annually, including Federal Emergency Manage-
ment Agency (FEMA) US&R teams. None of the spaces are
specifically designed for robots. Prop 133, shown in Fig. 1
is an exemplar of realistic deconstructed human habitable or
human restricted maneuverability indoor office building and
thus is a projection of what a SUAS will encounter. Portions
of the prop follow the floor plan and room size of a multi-
story government office building, such as the standing portions
of the Alfred P. Murrah Federal Building destroyed at the
Oklahoma City bombing in 1995. The prop consists of six
office-sized rooms on two floors, with four of the rooms
structurally intact and two rooms part of a lean-to collapse.
(a)
(b)
(c) (d)
Fig. 2: Interiors of Prop 133 at Disaster City R
. a) Furniture up
to heights of 1m to 2.5m scattered around the floor, b) Wires,
open ventilators, metal frames hanging from the ceiling at 2m
to 3m, c) Collapsed ceiling and wall, and d) Accumulated
debris due to breaking of loose material.
It should be noted that Prop 133 does not have the carpets,
wallpaper, acoustic tiles, or other organic materials normally
found in an office building, as those furnishings will mold in
the outdoors; therefore, Prop 133 may be less challenging for
robotic navigation and sensing than the partial collapse of an
actual office building.
II. REL ATE D WORK
Twelve systems were identified that have experimented with
indoor SUASs and four systems were intended for application
to search and rescue [6], [12], [1], [9]. The twelve systems
were evaluated in one or more of three testbeds: computer
simulation [12], [13], [14], [15], [11], physical - general indoor
environments [6], [9], [5], [7], [8], [10], [13], [14], [15], or
in a building that had experienced an actual disaster [1]. A
summary of the specifications of the twelve robots, the scale of
the testbed, the type of testbed, the type of regions represented
in the testbed, and the nominal flight altitude is provided in
Table I.
A. Computer Simulation
Five SUASs were evaluated using computer simulated
testbeds. Two of the five simulated environments emulated
habitable scale open spaces. Jongho and Youdan simulated
open spaces contained parallelepipeds, with sides of 1m and
3m [11]. Stowers et al. used a large number of blocks with
heights up to 3m [15], while Al Redwan Newaz et al. sim-
ulated an office space containing multiple objects positioned
on the floor with varying heights [12]. Two others simulated
a habitable scale office space. Fossel et al. simulated an office
space containing orthogonal walls, an open space containing
vertical pillars, and a collection of laboratory and office
space with a tilted wall [14]. The fifth system simulated a
restricted maneuverability office space [13]. The environments
represented randomly generated areas ranging in size from (25
x 25 x 3)m to (50 x 50 x 3)m. 20% of the environmental area
contained floor to ceiling walls and randomly placed obstacles,
such as boxes projecting from the floor up to a random height,
and fixed width beams mounted at random heights.
B. Physical - General Indoor Environment
Nine systems were evaluated in a physical - general indoor
testbed, with eight systems flying in habitable scale spaces and
only one in a restricted maneuverability scale space.
Eight of the nine systems were evaluated in habitable scale
environments [6], [9], [5], [7], [8], [10], [14], [15], while the
remaining system was evaluated in a restricted maneuverability
scale environment. Four of the environments are classified as
open space [9], [5], [7], [10]. Masanori et al.’s open space
environment contained three 1m cylinders and a horizontal
cross section that protruded from one cylinder at a height of
0.8m from the ground [5]. Torantani et al. operated in a testbed
with a rectangular obstacle 1m wide and 0.8m high, right at
the nominal flight altitude [7]. Suzuki et al.’s SUAS flew at
a nominal altitude of 1.5m, while avoiding a whiteboard of
similar height [9]. Ahrens et al.’s open space testbed contained
a 1m long cylindrical pole and a cuboid, similar to a bar
stool [10]. The SUAS flew at a nominal altitude of 0.5m,
as inferred from the paper, but flew up to 1.5m to avoid the
obstacles. The second most common habitable scale evaluation
testbed was an office space [6], [8], [14]. Li et al.’s SUAS
planned paths to allow the vehicle to avoid tables by flying
underneath them from one side to the other [6], while Fossel
et al.’s environment contained “low lying” tables, cabinets,
and benches [14]. Grzonka et al.’s SUAS flew at a maximum
of 1.5m, while avoiding 48cm high chairs and 77cm high
tables with other obstacles [8]. Stowers et al.’s laboratory
environment contained two large benches with instruments on
them, for a total height of 3m above the floor [15]. Their
SUAS flew at a nominal altitude of 1.5m, as inferred from
the paper. A 41m hallway provided a second environment in
which Grzonka et al.’s SUAS flew at a nominal altitude of
0.5m, as inferred from the paper.
MacAllister et al.’s [13] SUAS was the only system evalu-
ated for a restricted maneuverability scale environment. Their
environment was a collection of hallways, offices and open
spaces containing obstacles of random heights that were placed
on the floor and a horizontal bar placed at 0.7m above the
ground.
C. Search and Rescue
Four systems explicitly discussed the search and rescue
applications. Michael et al. [1] conducted evaluations in a
building on Tohoku University’s campus containing hallways
and offices. The building had been damaged by an earthquake,
but was still accessible to humans and was at the habitable
scale. Al Redwan Newaz et al. [12] considered a surveillance
and recovery mission after nuclear disasters or severe accidents
in industrial areas as an application, and tested a SUAS in
computer simulated habitable office space. Two systems [6],
[9] were evaluated in physical - general indoor staged office
and open spaces, respectively.
III. ENVIRONMENTAL CHARACTERISTICS
The environmental characteristics influencing the navigabil-
ity of a region can be divided into three groups: the scale
and degree of deconstruction, which captures the state of
the structure; the severity of obstacles and tortuosity, which
captures the impact of the deconstructed structure and dam-
aged furnishings; and other characteristics that affect sensing.
Table I provides an overview of the twelve surveyed SUASs
and the environmental characteristics for which they were
evaluated.
A. Scale and Degree of Deconstruction
The scale of the average size of the twelve SUASs, with
respect to the interior of Prop 133 is in the restricted maneuver-
ability range. If the floor plan is used to compute the Ecd, the
scale represents the habitable range, as Ecd >2Acd, where the
room is 6m and the SUAS is 0.6m, thus 6>2(0.6). However,
Figs. 2a and 2b show that the damage to fixtures and furnish-
ings reduce the actual free space to Ecd ≈1.5Acd, which falls
into the restricted maneuverability range of Ecd <2Acd.
As seen in the scale column of Table I, only two of the
twelve systems were deployed in a restricted maneuverability
(2Acd > Ecd >1.5Acd) environment, comparable to Prop
133. The remaining systems were evaluated or deployed in
environments within the habitable scale.
Prop 133 also illustrates different degrees of deconstruction
to the structural elements. Rooms 1 and 2 on both floors, as
seen in Figs. 2b and 2d have relatively minor deconstruction,
given that the walls, ceiling, and floor are still orthogonal,
though they exhibit holes or damage. Fig. 2c shows major de-
construction, where a ceiling has collapsed and the supporting
pillars are clearly damaged and no longer uniform.
Only one of the twelve SUASs was evaluated in a decon-
structed environment. Michael et al. deployed in a damaged
building [1], but with only with a minor degree of deconstruc-
tion compared to Prop 133. The other three systems proposed
for search and rescue missions [6], [12], [9] flew in regions
with no visible deconstruction. The remaining eight general
indoor SUASs operated in regions with no damage.
B. Severity of Obstacles and Tortuosity
Severity refers to the content of the environment, i.e.,
the number and types of obstacles an agent may encounter.
However, what is an obstacle for a human or a ground robot,
may not be an obstacle for a SUAS. Therefore, this paper rates
severity based on location:
1) Obstacles on the ground, below the nominal flying zone
(see Fig. 2a).
2) Obstacles on the ground, up to the nominal flying zone
(see Fig. 2a).
3) Obstacles hanging from the ceiling, in the nominal flying
zone (see Figs. 2a and 2b).
4) Obstacles hanging from the ceiling, above the nominal
flying zone (see Figs. 2a and 2b).
All four categories of obstacle severity/locations exist in the
environment at Prop 133. Altitude does not necessarily reduce
the obstacles. If a SUAS were to fly in Prop 133 at an altitude
of 1.17m (the average for flying in offices and hallways), it
will encounter the same categories of obstacles if it flew at
2.08m (the average for in open spaces).
Only one surveyed system, MacAllister et al. [13], was
evaluated in a testbed encompassing all four categories of
obstacle severity/location as found at Prop 133, but only in
computer simulation. The actual disaster deployment, Michael
et al. [1], encountered three types of obstacles, but not those
hanging from the ceiling into nominal flying zone. Al Redwan
Newaz et al. [12] simulated two categories of obstacles, while
the other two systems [6], [9] only tested with obstacles on
the ground up to the nominal flying zone. This observation
suggests that the obstacle placement in testbeds is not a good
predictor of whether a SUAS will be able to fly indoors during
a disaster.
Tortuosity is calculated as the number of turns taken by
the SUAS per unit distance. For example, if the SUAS takes
5 turns to avoid obstacles over a linear distance of 10m,
then the tortuosity is 5/10 = 0.5. The tortuosity at Prop
133 is estimated to be 1.0, i.e., 1 turn per meter. A low
tortuosity indicates that the frequency of obstacle avoidance is
low and the environment is comparatively easier to navigate,
as opposed to one with higher tortuosity.
All three types of spaces in Table II have a tortuosity much
lower than the tortuosity of Prop 133. The maximum tortuosity
in computer simulation (0.5), physical - general indoor staged
testbeds (0.31), physical - general indoor natural testbeds (0.1),
and actual disasters (0.6), suggests that the evaluation testbeds
are not sufficiently representative of actual disasters. A SUAS
that performs well in these testbeds may not have the agility
to make a higher frequency of turns and altitude changes.
C. Other Environmental Characteristics
Scale, degree of deconstruction, severity of obstacles, and
tortuosity are not the only environmental characteristics that
impact robot operations. Lighting conditions and surface
properties are two characteristics that have been observed at
disasters [4]. The conditions are not adequately captured in
the photographs of Prop 133, but merit further discussion. The
Prop 133 exhibited natural, non-uniform and low light condi-
tions, due to collapsed ceilings and holes in walls, shadows
due to obstacles, and lack of windows. Cameras do not work
No. Author SUAS Diameter (m) Scale Testbed Region Nominal Altitude (m)
1 Masanori et al., 2013 – Habitable Physical - General
Indoor (Staged)
Open Space 0.7
2 Li et al., 2013 0.57 Habitable Physical - General
Indoor (Staged)
Office –
3 MacAllister et al., 2013 – Restricted Man. Physical - General
Indoor (Staged)
Collection of hallways,
offices and open space
0.7
Computer Office –
4 Jongho et al., 2013 – Habitable Computer Open Space 7.0
2.0
5 Al Redwan Newaz et al., 2013 – Habitable Computer Office –
6 Fossel et al., 2013
0.73
Habitable
Physical - General
Indoor (Natural)
Office –
– Computer
Collection of lab and
office
–
Open Space –
Lab –
7 Toratani et al., 2013 0.54 Habitable Physical - General
Indoor (Staged)
Open Space 0.8
8 Grzonka et al., 2013 – Habitable Physical - General
Indoor (Natural)
Hallway 0.5
Office –
9 Michael et al., 2012 0.65 Restricted Man. Actual Disaster
Environment
Collection of Offices
and Hallways
2.0
10 Stowers et al., 2011 – Habitable Physical - General
Indoor (Staged)
Lab 1.5
Computer Open Space –
11 Suzuki et al., 2010 1.0 Habitable Physical - General
Indoor (Staged)
Open Space 1.5
12 Ahrens et al., 2009 0.54 Habitable Physical - General
Indoor (Staged)
Open Space 0.5
TABLE I: A summary of the reviewed systems, including SUAS size, environmental scale, space classification and nominal
altitude. (– means “unable to be inferred”).
well in dim lighting and may need an artificial light source,
while the Kinect does not work in high luminesce conditions,
as shown in [16]. The building materials for Prop 133 include
metal, glass and sharp edges that may scatter active sensors,
such as LIDAR and Ultrasound. Carpets, cloth, soundproof
tiles and partitions typically found in office buildings may
absorb sound signals. Furthermore, suspended dust due to
debris and loose building materials may affect the visibility
and make sensors less effective or even non-functional.
IV. CONCLUSIONS
The interest in using SUAS technology by urban search
and rescue teams continues to grow; however, the unique
situations in which SUASs will be considered a valuable
tool place constraints on the system and algorithm design.
This paper focuses on providing formal definitions of the
environmental constraints based on extending definitions for
disaster response unmanned ground systems. These definitions
are critical for purposes of developing and evaluating SUAS
technology for and within representative environments that
will lead to transferring the technology to disaster response
personnel. Kinetic disasters modify building structures and
contents, thus requiring a SUAS to meet very different con-
straints for successful deployment.
The formal definitions characterize the environment’s scale,
space type, obstacle severity and tortuosity as well as the
SUAS’ nominal flight altitude. These definitions were used
to analyze and classify twelve existing SUASs evaluated for
deployment in simulation or in actual indoor environments.
These results were compared to a representative environment
used to train urban search and rescue teams, Prop 133 at
Disaster City R
. While one of the twelve systems surveyed
was deployed in a building damaged by an earthquake, and
three other systems claimed applicability to search and rescue
domains, none of the systems were evaluated in environments
representative of Prop 133. While it is not possible to say
with certainty that the analyzed SUASs cannot fly in the
Prop 133 region, it is clear that their obstacle avoidance and
simultaneous localization and mapping algorithms have not
been evaluated for such environments. It is also likely, al-
though this paper does not analyze this aspect, that the sensors
supporting these algorithms will likely produce inaccuracies
for environments represented by Prop 133. As such, it is
necessary to consider the environmental characteristics defined
in this paper when designing the hardware and software
specifications for new SUASs intended for indoor urban search
and rescue environments.
V. ACKNOW LED GEM ENT S
Portions of this work were supported by a SEC Travel
grant and NSF Grant IIS-1143713 EAGER: Shared Visual
Common Ground in Human-Robot Interaction for Small Un-
manned Aerial Systems. The authors thank the Texas A&M
Engineering Extension Service for access to their facilities.
No Author from Table I Testbed Severity: Obstacle Location Tortuosity
Ground
to below
nominal
Ground
up to
nominal
Ceiling
into
nominal
Ceiling
to above
nominal
1 Masanori et al., 2013 Physical - General
Indoor (Staged)
X0.5
2 Li et al., 2013 Physical - General
Indoor (Staged)
X–
3 MacAllister et al., 2013 Physical - General
Indoor (Staged)
XXX 0.18
Computer XXXX0.4
4 Jongho et al., 2013 Computer X0.14
X0.2
5 Al Redwan Newaz et al., 2013 Computer
X X –
X X –
X X –
6 Fossel et al., 2013
Physical - General
Indoor (Natural)
X X –
Computer
X X –
X–
X X –
7 Toratani et al., 2013 Physical - General
Indoor (Staged)
X0.3
8 Grzonka et al., 2013 Physical - General
Indoor (Natural)
X X 0.1
X X –
9 Michael et al., 2012 Actual Disaster
Environment
XX X0.6
10 Stowers et al., 2010 Physical - General
Indoor (Staged)
X X –
Computer X X 0.57
11 Suzuki et al., 2010 Physical - General
Indoor (Staged)
X–
12 Ahrens et al., 2009 Physical - General
Indoor (Staged)
X0.5
TABLE II: Summary of Severity of Obstacles and Tortuosity. (– means “unable to be inferred”)
REFERENCES
[1] N. Michael, S. Shen, K. Mohta, Y. Mulgaonkar, V. Kumar,
K. Nagatani, Y. Okada, S. Kiribayashi, K. Otake, K. Yoshida,
K. Ohno, E. Takeuchi, and S. Tadokoro, “Collaborative mapping of an
earthquake-damaged building via ground and aerial robots,” Journal of
Field Robotics, vol. 29, no. 5, pp. 832–841, 2012. [Online]. Available:
http://dx.doi.org/10.1002/rob.21436
[2] G.-J. Kruijff, V. Tretyakov, T. Linder, F. Pirri, M. Gianni, P. Papadakis,
M. Pizzoli, A. Sinha, E. Pianese, S. Corrao, F. Priori, S. Febrini, and
S. Angeletti, “Rescue robots at earthquake-hit mirandola, italy: a field
report,” in IEEE International Symposium on Safety, Security and Rescue
Robotics, pp. 1–8.
[3] K. Pratt, R. Murphy, S. Stover, and C. Griffin, “Conops and autonomy
rcommendations for vtol suass based on hurricane katrina operations,”
Journal of Field Robotics, vol. 26, no. 8, pp. 636–650, 2009.
[4] R. R. Murphy, Disaster Robotics. MIT Press, 2014.
[5] H. Masanori, N. Hideyuki, S. Johan, and B. Kevin, Optimal
Trajectory Generation and Tracking Control of a Single Coaxial
Rotor UAV, ser. Guidance, Navigation, and Control and Co-
located Conferences. American Institute of Aeronautics and
Astronautics, 2013, doi:10.2514/6.2013-4531. [Online]. Available:
http://dx.doi.org/10.2514/6.2013-4531
[6] Q. Li, D.-C. Li, Q.-f. Wu, L.-w. Tang, Y. Huo, Y.-x. Zhang,
and N. Cheng, “Autonomous navigation and environment modeling
for mavs in 3-d enclosed industrial environments,” Computers in
Industry, vol. 64, no. 9, pp. 1161–1177, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0166361513001243
[7] D. Toratani, T. Higuchi, and S. Ueno, “Terrain following flight of
uav using information amount feedback,” in SICE Annual Conference
(SICE), 2013 Proceedings of, Conference Proceedings, pp. 1503–1508.
[8] S. Grzonka, G. Grisetti, and W. Burgard, “A fully autonomous indoor
quadrotor,” Robotics, IEEE Transactions on, vol. 28, no. 1, pp. 90–100,
2012.
[9] R. Suzuki, T. Matsumoto, A. Konno, Y. Hoshino, K. Go, A. Oosedo,
and M. Uchiyama, “Teleoperation of a tail-sitter vtol uav,” in Intelligent
Robots and Systems (IROS), 2010 IEEE/RSJ International Conference
on, Conference Proceedings, pp. 1618–1623.
[10] S. Ahrens, D. Levine, G. Andrews, and J. P. How, “Vision-based
guidance and control of a hovering vehicle in unknown, gps-denied
environments,” in Robotics and Automation, 2009. ICRA ’09. IEEE
International Conference on, Conference Proceedings, pp. 2643–2648.
[11] P. Jongho and K. Youdan, Obstacle Detection and Collision Avoidance
of Quadrotor UAV Using Depth Map of Stereo Vision, ser. Guidance,
Navigation, and Control and Co-located Conferences. American
Institute of Aeronautics and Astronautics, 2013, doi:10.2514/6.2013-
4994. [Online]. Available: http://dx.doi.org/10.2514/6.2013-4994
[12] A. Al Redwan Newaz, F. A. Pratama, and C. Nak Young, “Exploration
priority based heuristic approach to uav path planning,” in RO-MAN,
2013 IEEE, Conference Proceedings, pp. 521–526.
[13] B. MacAllister, J. Butzke, A. Kushleyev, H. Pandey, and M. Likhachev,
“Path planning for non-circular micro aerial vehicles in constrained envi-
ronments,” in Robotics and Automation (ICRA), 2013 IEEE International
Conference on, Conference Proceedings, pp. 3933–3940.
[14] J. Fossel, D. Hennes, D. Claes, S. Alers, and K. Tuyls, “Octoslam: A 3d
mapping approach to situational awareness of unmanned aerial vehicles,”
in Unmanned Aircraft Systems (ICUAS), 2013 International Conference
on, Conference Proceedings, pp. 179–188.
[15] J. Stowers, M. Hayes, and A. Bainbridge-Smith, “Beyond optical flow
- biomimetic uav altitude control using horizontal edge information,” in
Automation, Robotics and Applications (ICARA), 2011 5th International
Conference on, Conference Proceedings, pp. 372–377.
[16] J. Suarez and R. Murphy, “Using the kinect for search and rescue
robotics,” in Safety, Security, and Rescue Robotics (SSRR), 2012 IEEE
International Symposium on, Nov 2012, pp. 1–2.
