Toward an Accurate Hydrologic Urban Flooding Simulations for

Disaster Robotics

Marcelo Paravisi

1,2

, Vitor A. M. Jorge

and Alexandre Amory

Instituto Federal de Educac¸

ao, Ci

encia e Tecnologia do Rio Grande do Sul, Brazil

School of Technology, PUCRS University, Brazil

Keywords:

Robotics Simulation, Disaster Robotics, Disaster Simulation, Urban Flooding, Hydrological Models, USV.

Abstract:

Testing and benchmarking robots in actual disaster scenarios is risky and sometimes nearly impossible. The

lack of adequate tests could make robots more vulnerable and less effective in an actual disaster situation.

However, even if it was possible to test them in a disaster scenario, the test itself would have high risks for

the equipment and the robot operator. For this reason, simulations can be a powerful alternative to validate

unmanned systems in safe and controlled virtual environments. The main challenge is to devise an accurate

virtual scenario as close as possible to an actual disaster scenario. This problem is particularly harder if the

robot in question is an unmanned surface vehicle (USV), mainly due to the numerous disturbances which can

affect the robot. This paper presents and discusses the simulation of an urban ﬂooding scenario with accurate

environmental disturbances faced by an USV, such as water currents, waves and winds. Results demonstrate

that these environmental disturbances have a relevant effect on the USV’s ability to perform basic navigational

tasks. The main conclusion of this work is that there is a long road ahead of USV simulators in order to

validate USVs in realistic disaster scenario simulations.

1 MOTIVATION

Extreme events have terrible effects wherever they oc-

cur. Be it a natural disaster or one caused by men, the

result is often loss of lives, injured people, as well

as the destruction of both the individual possessions

(houses, cars, and so on) and the basic utilities infras-

tructure (water supply, electricity, communications,

and hospitals). The result is often homeless people

with all sort of basic needs with their lives at risk not

only due to their injuries, but also due to hazards im-

posed by the new harsh and dangerous post-disaster

environment, where ﬁrst responders often must put

their lives at stake to reach and rescue others.

Thus, unmanned systems can be a valuable tool

for disaster responders by going to places which

otherwise would be too dangerous for ﬁrst respon-

ders. As such, rescue robots typically face dull, dirty,

and harsh environments with poor infrastructure (e.g.

blocked streets, destroyed buildings, poor connectiv-

ity, dust, extreme heat, and so on) (Murphy, 2014)

and harsh meteorological conditions, such as strong

winds, storms, river overﬂows, large waves, and so

on.

Back in the early 2000’s, robots started to be de-

ployed for disaster response (Murphy, 2012) and soon

it became clear that disaster robotics was a ﬁeld on

its own. Since then, they were used in several dis-

aster sites, including collapsed mines & buildings,

earthquakes, tornadoes, landslides, and ﬂoods (Mur-

phy, 2012; Murphy, 2014). While some degree of

success was achieved, conventional robot design has

proven to lack robustness, mainly due to extreme

harsh conditions of disaster sites. In 2004, a study

on robots used for rescue in urban areas, encompass-

ing 15 robots from three different manufacturers, has

shown that the Mean Time Between Failures (MTBF)

of rescue robots was 24 hours while their availabil-

ity was around 54% (Carlson et al., 2004). The study

of disaster robotics was on its infancy, but the impor-

tance of robots for risky interventions soon became

clear (Habib and Baudoin, 2010), along with the ex-

treme challenges faced by robots in harsh environ-

ments (Habib and Baudoin, 2010; Wong et al., 2017).

Another challenge preventing the success of dis-

aster robots may be the difﬁculty to evaluate them

properly, since there are very few places to test them

which are similar to actual disaster sites. Without

proper testing, the robot will likely fail during the

actual use in a disaster. The importance of evalu-

Paravisi, M., Jorge, V. and Amory, A.

Toward an Accurate Hydrologic Urban Flooding Simulations for Disaster Robotics.

DOI: 10.5220/0006904704250431

In Proceedings of the 15th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2018) - Volume 2, pages 425-431

ISBN: 978-989-758-321-6

425

Table 1: UXV Disturbances’ Inﬂuence.

Unmanned System Winds Water Current Waves Nearby UXV

UGV weak — — weak

UAV strong — — strong (close by)

UUV — strong strong (surface) moderate (close by)

USV moderate strong strong moderate

ation and benchmarking for disaster robots in con-

trollable conditions (Murphy et al., 2008) soon be-

came a recognized research problem. Initiatives such

as the Robocup-Rescue Project (Takahashi and Ta-

dokoro, 2002) and League (Balakirsky et al., 2007)

brought the problem to the spotlight. Actual dis-

aster emulation sites, such as the National Insti-

tute of Standards and Technology (NIST)’s Disas-

ter City®(Disaster City encompasses a 52 acre real-

world benchmarking environment (Khoury and Ka-

mat, 2009; Wilde et al., 2015).), were designed for

testing disaster robots in extreme conditions. Compe-

titions focusing on marine robotics, such as SAUC-

E (http://sauc-europe.org/), featuring AUVs, and eu-

Rathlon (http://www.eurathlon.eu/), featuring under-

water, surface and aerial robots, were designed to test

unmanned systems in harsh water environments.

Given such challenging environments and the

risks for costly robot platforms and humans, sim-

ulations could come as important assets for reduc-

ing costs of disaster robotics research, by detecting

early problems in robot design and avoiding epic mis-

sion failures when human lives are at stake. Even

though there is little argument that simulations are

easier than real life scenarios (Murphy et al., 2008),

when it comes to water environments and unmanned

surface vehicles (USVs), the number of freely avail-

able simulators is small when compared to other plat-

forms (Torres-Torriti et al., 2016). In fact, when we

talk about realistic simulations of harsh water envi-

ronments this number drops to zero. All that, given

the risks associated with water disturbances and the

importance of USVs.

USVs are valuable assets for disaster missions in-

volving maritime or ﬂooded environments. They can

serve as emergency data/communication relays, or

the source of real-time video from disaster sites. In

addition, they can perform underwater assessments

(through bathymetry), detecting debris, accretion or

erosion caused by ﬂoods transports. Furthermore,

they can serve as emergency carriers of 1st aid kits,

potable water, or food. Finally, they can even serve as

docking stations and communications base for other

unmanned systems. Compared to other platforms,

however, USVs are the ones which are inﬂuenced by

the greater number of environmental factors: winds,

waves; and wind currents. In addition, they ﬂoat over

the water surface, which makes USVs easily inﬂu-

enced by other USVs and manned embarkations, even

if they are not close to one another, making it chal-

lenging to simulate the operation of multiple vehicles

simultaneously. Table 1 compares the inﬂuence of im-

portant environmental disturbances for UGVs, UAVs,

UUVs and USVs control.

We collected information about ten open source

simulators for marine robots available in the literature

(Kelpie (Mendona et al., 2013), USARSim (Nour-

bakhsh et al., 2005; Sehgal and Cernea, 2010),

MARS (Tosik and Maehle, 2014), Stage (Gerkey

et al., 2003), MOOS-IVP (Benjamin et al., 2010),

UW-Morse (Henriksen et al., 2016), UWSim (Prats

et al., 2012), the Gazebo (Koenig and Howard, 2004),

“FreeFloating” plugin (Kermorgant, 2014), V-REP

(Rohmer et al., 2013)). The number of issues in-

volving currently available simulators include: lack

of code availability (Kelpie, UW-Morse); discontin-

ued software (MARS); and outdated libraries (US-

ARSim). In addition, some simulators which do not

lack setup problems, still have limited to none built-in

simulation of disturbances (Stage and MOOS-IVP).

The remaining simulators do have some environmen-

tal disturbances simulation capabilities, however none

of them are ready to simulate harsh conditions for

USVs. Besides, they do not even make it easy to

spawn USVs, lacking a ready-to-use USV examples

– i.e., USVs must be designed from scratch includ-

ing model and control features. From Table 1, we

see that disturbances are rather important for USVs,

even though disturbances in available simulators are

limited. None include harsh wave conditions, usually

handling calm waters without turbulence. Besides,

current wind and water disturbances simulations ig-

nore the inﬂuence of the actual terrain relief, pre-

venting physically and hydrologically accurate sim-

ulations – e.g., strong winds, river currents, turbulent

waters, large waves and so on. All that makes it difﬁ-

cult to validate simulations of USVs in actual disaster

sites.

In this paper, we advocate for improved simula-

tions for USVs, both in terms of disturbances and

sensor noise to enable minimum proper validation of

USV systems. The goal of this work is to discuss the

ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics

426

wate

current

yaml

file

wind

current

yaml

file

velocity

Ros Topic: time

GAZEBO

UWSIM

POSE

WAVE

HEIGHT

Figure 1: Overall simulation architecture and main mod-

ules. The blue modules are new or improved.

creation of a simulated ﬂooding scenario of an actual

urban environment (Porto Alegre, Brazil) and imple-

ment an initial version in a robotic simulator. Given a

Digital Elevation Model (DSM) of the terrain, hydro-

logical and wind models are generated to accurately

reproduce the environmental forces (wind and water

current) of a large scale ﬂooded site. We present ini-

tial results while evaluating an USV, where experi-

ments show that such disturbances have a strong ef-

fect on USV behavior. This work also discusses the

numerous gaps and requirements for current and fu-

ture open-source disaster simulators.

This paper is presented as follows. Section 2

presents the simulation environment. Section 3 de-

scribes the construction of the urban scenario, while

Section 4 shows a case study of a USV crossing

a ﬂooded environment considering winds and water

disturbances. Section 5 concludes the paper with a

discussion and future work.

2 SIMULATION CAPABILITIES

The present work is based on the USVSim simulator,

along with its modules, and the scenarios (freely

available for download at https://github.com/disaster-

robotics-proalertas/usv

sim lsa). This section

presents an overview of the architecture of the

simulator (see Fig. 1) — the blue boxes represent

the new or customized simulation modules. The

complete description is presented in (Paravisi et al.,

2018).

The simulator works a series of Gazebo modules

which enable it to simulate environment disturbances

for USVs, while UWSim is only used for visualiza-

tion purposes due to its improved water rendering

capability compared to Gazebo. As represented in

Fig. 1, the core of Gazebo is not modiﬁed but a new

modules are included, such as the usv sailing plugin.

The USV Sailing plugin can simulate the forces of

(a)

(b)

Figure 2: The black dots represent the center of buoyancy,

while red and blue arrows represent, respectively, the grav-

ity and buoyancy force vectors. In (a), the resulting buoy-

ancy effect of the boat when represented by a set of links

and joints (b). Note that the gravity and buoyancy forces

are applied to each of the links’ center of buoyancy.

water current over the boat rudder. In order to es-

timate those forces, it reads the true water velocity

from usv water plugin. After that, it computes the

forces applied to the foil of the rudder, considering

the speed of the ﬂuid and the foil to compute the lift

and drag forces.

When a vehicle pose is updated in UWSim, the

wave height relative to the vehicle’s center(s) of buoy-

ancy is sent to the Improved Freeﬂoating Gazebo

to compute the buoyancy effect. The Improved

Freeﬂoating Gazebo plugin allows USVs to suffer roll

and pitch orientation changes caused by wave motion.

In such strategy, the USV hull must be subdivided into

several parts, i.e. modeled by a set of links bounded

together by ﬁxed joints (see Fig. 2b). The resulting

effect of the gravity (red arrow) and buoyancy (blue

arrow) forces is depicted in Fig. 2a, allowing the boat

to roll and pitch more naturally, following the shape

of the wave.

The water and wind current generators are mod-

eled as ROS nodes which receive requests from Im-

proved FreeFloating Gazebo to enhance the boat mo-

tion realism through wind and current information.

Besides, the water current generator is used by the

usv sailing plugin to compute the force which is di-

rectly applied to the boat, using as input the velocities

of the boat and of the water/wind currents.

The usv water current module loads data exported

from the HEC RAS (W. Brunner, 1995) hydrological

simulator. Thus, users can simulate the ﬂow of rivers

by inserting simple height maps, then exporting Hi-

erarchical Data Format (HDF) ﬁles, which store the

water velocities for each time step of the simulated

water ﬂow. Then, the simulation architecture requests

the velocity of the water at each of the USV’s posi-

tion of links. Similarly, the usv wind current module

Toward an Accurate Hydrologic Urban Flooding Simulations for Disaster Robotics

427

Qgis manipulation

Hydrologic Simulation

(HEC-RAS)

USV_Sim

(Robotic Simulator)

USV_WATER_CURRENT USV_WIND_CURRENT

Digital Surface Model

Filtered

Digital Surface Model

Output

Simulation

(HDF file)

Parameter

Configuration

Yaml File

Scenario

Description

(xml)

Wind

Collision

map

Parameter Configuration

Yaml File

Figure 3: Pre-processing steps to generate hydrological and

wind current maps.

simulates the wind over a 2D region with of multiple

obstacles using a method based on the ﬂuid dynam-

ics module with the Lattice-Boltzmann method (Qian

et al., 1992). The wind force affects the part of the

hull above the line of water. The wind collision map

is extracted from the scenario.

3 MODELING A DISASTER

SCENARIO

Fig. 3 shows the steps required to build the hydro-

logical and wind current maps for a given 3D sce-

nario. To simulate accordingly the ﬂow of rivers

with hydrological models, we should consider the di-

rect impact the topology of terrain has in the ﬂow

of water. Thus, the ﬁrst step is to acquire a Digi-

tal Surface Model (DSM), i.e. a model that incor-

porates the topology of the terrain and the shape and

heights of buildings. There are some on-line reposi-

tories that allows the user download this type of data

(https://earthexplorer.usgs.gov/), but in some cases

the DSM does not include bathymetry (underwater el-

evation) information, which can be useful to identify

safe routes for boats and USVs.

Eventually, some features of the DSM may have

to be ignored (e.g. small trees given the high resolu-

tion of the DSM), therefore, the user can edit and ﬁlter

out the undesired features. We accomplish that using

the QGIS(https://qgis.org/en/site/). open-source geo-

graphic information system.

Then, the resulting DSM is used as input to the

HEC-RAS hydrological simulator. In order to simu-

late water ﬂowing through the scenario, we have to

deﬁne a 2D grid over the terrain and the location of

the upper and lower river reaches. After that, the user

deﬁnes the initial water level conditions and, to all

river reaches, the user can associate the water volume

) at each simulation time step. This conﬁguration

allows the variation of water ﬂow over time. The ﬁ-

nal step is to run the simulation and export an HDF

ﬁle, which includes the velocity ﬁeld for each grid cell

over time for the entire simulation of the water ﬂow.

This HDF ﬁle deﬁned in a conﬁguration ﬁle (YAML

ﬁle) is used as input to the usv water current module.

For the simulation of winds, the user should cre-

ate a conﬁguration ﬁle (YAML extension), inform-

ing the name of ﬁltered DSM, the wind direction and

speed, grid resolution and the ﬁlename of an user-

deﬁned collision map. This map is a bitmap with the

same resolution as the ﬁltered DSM, allowing the user

to manually add static obstacles which interfere with

wind ﬂow but which were not present in DSM at the

time of its creation – e.g. buildings or a large ship.

This way, the user can personalize the wind collision

map using only an image editor, without the need to

resort to DSM or QGIS manual edits.

The top view of the resulting ﬂooding scenario can

be seen in Fig. 4a. The water ﬂow comes from the

right-hand side of the image, at the location of the Dil-

vio river, in Porto Alegre, Brazil. The simulated area

covers an area of 430 meters by 400 meters. Fig. 4b

shows the direction of the water ﬂow and the water

velocity in the ﬂooded scenario, the whiter the color,

the deepest the channel. While the particles trajec-

tories show the water current speed, so longest parti-

cles means that water current is fastest. One can see

that the parts close to the original river bed have faster

water currents than those nearby the buildings. Also,

Fig. 4c shows the wind map of the same region, where

heat map indicates the intensity of wind.

Fig. 5 shows another screenshot from the modeled

scenario, but from a point of the view right behind the

boat, showing the ﬂooded buildings, trees, and infras-

tructure.

4 PRELIMINARY RESULTS

The implemented Guidance, Navigation and Control

(GNC) strategy is based on a built-in heading PID

controller, allowing them to reach desired positions in

the environment. This control strategy computes the

angle between the boat position and target(waypoint)

position. With this angle, the position of actuators are

updated to robot follow the desired orientation.

Fig. 6 shows the different trajectories when the

ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics

428

(a) Aerial photo from simulated scenario.

(b) ﬂooded scenario and water current intensity

Figure 4: Aerial view of the testing site in Google Maps

(a), the water current (b) and wind intensity maps (c) for

the ﬂooded scenario.

Figure 5: Simulation of the USV at a ﬂooded site.

Figure 6: USV trajectories with (red) and without (blue)

water & wind currents.

Table 2: Total travel time and distance to destination.

Disturbances Time (s) Distance (m)

no 141 157.0

yes 148 167.6

water/wind currents are on and off. Several equally

spaced waypoints where inserted along the path of

160 meters. One can see that when the rudder boat

is in the fastest water current part the trajectory, error

increases considerably. On the other hand, the error

reduces as the boat reaches the margins, where the

water current is slower. However, the wind changes

velocity and orientation near by the buildings.

Tab. 2 shows the time to perform the path and the

total traveled distance.

Toward an Accurate Hydrologic Urban Flooding Simulations for Disaster Robotics

429

5 DISCUSSION & FUTURE

WORK

This paper presents a strategy to model realistic ﬂood-

ing scenarios in a robotic simulator in order to eval-

uate rescue USVs on harsh environments safely and

cheaply when compared to an actual ﬁeld test in a

disaster site. Initial experiments show the importance

of proper disturbance modeling and simulation in the

control of USVs, which may even prevent the USV

from achieving its goal or cause collisions — unac-

ceptable for disaster response missions.

Beyond the presented ﬂooding scenario, we plan

to simulate several other types of disasters, such as,

the effects of tsunamis or dam breaches. This would

allow new GNC algorithms to be evaluated even in

such harsh environment conditions.

However, there is much to be done to improve

simulations in maritime environments. Current open-

source simulators for USVs still lack many features.

Proper simulation of sensors and communications

problems with USVs in simulations are often limited

to Gaussian noise. Instead, intermittent and unreliable

services should be considered to validate the robust-

ness of the whole system. These could be achieved

by considering factors such as the simulation of wire-

less transmissions’ shadows and GPS errors close to

buildings and trees, as well as the effect of water dis-

turbances & weather on sensors and communications.

For instance, radio waves can be reﬂected by the water

surface and waves, sporadicly interrupting wireless

communications. The same is true for weather con-

ditions and heavy rain, which affects cameras and the

range of communications. Finally, underwater data

transmission, should also consider physically correct

signal attenuation and multi-path simulations. By do-

ing that, the validation of unmanned systems in simu-

lation environments will be more similar to that which

UAV, UGV and USVs facing actual disaster missions.

In order to improve the realism of disaster loca-

tions, debris could be added to scene, so they could

become ﬂoating obstacles which could even be car-

ried by the water ﬂow, colliding with USV and UUV.

If needed, users could combine those objects together

(by adding joints) and deﬁning a limit force that

would break their joints. This way, when the grouped

object collides with another one with considerable

strength, many parts or pieces can be detached gener-

ating even more debris deposited all over the city and

into water channels, ports and bays. Then, the result-

ing scenario can be used to emulate post disaster as-

sessment missions, where the amount and location of

debris must be estimated by an heterogeneous team of

unmanned systems. As a result, new strategies could

be designed and properly tested & compared for area

coverage, debris detection rate and their removal from

affected areas in realistic simulation environments.

Future works may also include the integration of

UAVs (affected by winds) and USVs (affected by

winds, waves and water currents) collaborating in the

same scenario where they look for strained people.

We also plan to model landslides, bigger turbulent

waves (tsunami like), bridge and oil platform col-

lapses to assess, in simulations, heterogeneous teams

of unmanned systems in such marine disaster environ-

ments.

ACKNOWLEDGEMENTS

This paper was partially funded by CAPES/Brazil,

under project 88887.115590/2015-01, Pro-Alertas

program.

The MDT and MDS maps belong to the munici-

pality of Porto Alegre, Brazil. The maps were pro-

vided by the Secretary of Municipal Urbanism (Sec-

retaria Municipal de Urbanismo - SMURB) to profes-

sor Regis Lahm.

REFERENCES

Balakirsky, S., Carpin, S., Kleiner, A., Lewis, M., Visser,

A., Wang, J., and Ziparo, V. A. (2007). Towards het-

erogeneous robot teams for disaster mitigation: Re-

sults and performance metrics from robocup rescue.

Journal of Field Robotics, 24(11-12):943–967.

Benjamin, M. R., Schmidt, H., Newman, P. M., and

Leonard, J. J. (2010). Nested autonomy for unmanned

marine vehicles with MOOS-IvP. J. Field Robotics,

27(6):834–875.

Carlson, J., Murphy, R. R., and Nelson, A. (2004). Follow-

up analysis of mobile robot failures. In IEEE In-

ternational Conference on Robotics and Automation

(ICRA), volume 5, pages 4987–4994.

Gerkey, B. P., Vaughan, R. T., and Howard, A. (2003). The

Player/Stage project: Tools for multi-robot and dis-

tributed sensor systems. In International Conference

on Advanced Robotics (ICRA), pages 317–323.

Habib, M. K. and Baudoin, Y. (2010). Robot-assisted

risky intervention, search, rescue and environmen-

tal surveillance. International Journal of Advanced

Robotic Systems, 7(1):10.

Henriksen, E. H., Schjlberg, I., and Gjersvik, T. B. (2016).

UW-MORSE: The underwater modular open robot

simulation engine. In IEEE/OES Autonomous Under-

water Vehicles (AUV), pages 261–267.

Kermorgant, O. (2014). A dynamic simulator for underwa-

ter vehicle-manipulators. In Brugali, D., Broenink,

J. F., Kroeger, T., and MacDonald, B. A., editors,

International Conference on Simulation, Modeling,

and Programming for Autonomous Robots (SIMPAR),

ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics

430

volume 8810 of Lecture Notes in Computer Science,

pages 25–36. Springer.

Khoury, H. M. and Kamat, V. R. (2009). Evaluation of po-

sition tracking technologies for user localization in in-

door construction environments. Automation in Con-

struction, 18(4):444 – 457.

Koenig, N. and Howard, A. (2004). Design and use

paradigms for gazebo, an open-source multi-robot

simulator. In IEEE/RSJ International Conference

on Intelligent Robots and Systems (IROS), volume 3,

pages 2149–2154.

Mendona, R., Santana, P., Marques, F., Loureno, A., Silva,

J., and Barata, J. (2013). Kelpie: A ROS-based multi-

robot simulator for water surface and aerial vehicles.

In IEEE International Conference on Systems, Man,

and Cybernetics, pages 3645–3650.

Murphy, R. R. (2012). A decade of rescue robots. In

2012 IEEE/RSJ International Conference on Intelli-

gent Robots and Systems, pages 5448–5449.

Murphy, R. R. (2014). Disaster Robotics. The MIT Press.

Murphy, R. R., Tadokoro, S., Nardi, D., Jacoff, A., Fiorini,

P., Choset, H., and Erkmen, A. M. (2008). Search and

Rescue Robotics, pages 1151–1173. Springer Berlin

Heidelberg, Berlin, Heidelberg.

Nourbakhsh, I. R., Sycara, K., Koes, M., Yong, M., Lewis,

M., and Burion, S. (2005). Human-robot teaming

for search and rescue. IEEE Pervasive Computing,

4(1):72–79.

Paravisi, M., Santos, D. H., Jorge, V. A. M., Gonc¸alves,

L. M., and Amory, A. (2018). Unmanned surface

vehicle simulator with environmental disturbances.

In IEEE/RSJ International Conference on Intelligent

Robots and Systems (IROS), pages 1–7. submitted pa-

per.

Prats, M., Prez, J., Fernndez, J. J., and Sanz, P. J. (2012). An

open source tool for simulation and supervision of un-

derwater intervention missions. In IEEE/RSJ Interna-

tional Conference on Intelligent Robots and Systems

(IROS), pages 2577–2582.

Qian, Y. H., D’Humires, D., and Lallemand, P. (1992). Lat-

tice BGK models for navier-stokes equation. EPL

(Europhysics Letters), 17(6):479.

Rohmer, E., Singh, S. P. N., and Freese, M. (2013). V-REP:

A versatile and scalable robot simulation framework.

In IEEE/RSJ International Conference on Intelligent

Robots and Systems (IROS), pages 1321–1326.

Sehgal, A. and Cernea, D. (2010). A multi-AUV missions

simulation framework for the USARSim robotics sim-

ulator. In Mediterranean Conference on Control Au-

tomation (MED), pages 1188–1193.

Takahashi, T. and Tadokoro, S. (2002). Working with robots

in disasters. IEEE Robotics Automation Magazine,

9(3):34–39.

Torres-Torriti, M., Arredondo, T., and Castillo-Pizarro, P.

(2016). Survey and comparative study of free simula-

tion software for mobile robots. Robotica, 34(4):791–

822.

Tosik, T. and Maehle, E. (2014). MARS: A simulation en-

vironment for marine robotics. In 2014 Oceans - St.

John’s, pages 1–7.

W. Brunner, G. (1995). HEC-RAS river analysis system.

hydraulic reference manual. version 1.0. Technical re-

port, DTIC.

Wilde, G. A., Murphy, R. R., Shell, D. A., and Marianno,

C. M. (2015). A man-packable unmanned surface

vehicle for radiation localization and forensics. In

2015 IEEE International Symposium on Safety, Secu-

rity, and Rescue Robotics (SSRR), pages 1–6.

Wong, C., Yang, E., Yan, X. T., and Gu, D. (2017). An

overview of robotics and autonomous systems for

harsh environments. In 2017 23rd International Con-

ference on Automation and Computing (ICAC), pages

1–6.

Toward an Accurate Hydrologic Urban Flooding Simulations for Disaster Robotics

431