Robotic lighter-than-air vehicles, or aerobots, have astrategic potential as surveying and instrumentdeployment platforms for the exploration of planets andmoons with an atmosphere, such as Venus, Mars andTitan. Aerobots are characterized by modest powerrequirements, extended mission duration and longtraverse capabilities, as well as the ability to transportand deploy scientific instruments and in-situ laboratoryfacilities over vast distances. With the arrival of theHuygens probe at Saturn’s moon Titan in early 2005,there is considerable interest in a subsequent follow-onmission that would explore Titan’s surface through asubstantially autonomous aerobot. Autonomousoperation is required due to the nominal 2.6 hoursround trip communication delay between Earth andTitan, as well as multi-day communication blackoutscaused by Titan’s rotation and orbit around Saturn. Inthis paper, we discuss first steps towards thedevelopment of an autonomy architecture and a coreset of perception, reasoning and control technologiesfor a future Titan aerobot. We provide an overview ofthe design of the autonomy architecture, whichintegrates perception-based inferences about theenvironment of operation of the vehicle, vehicle healthmonitoring and reflexive safing actions, accurate flightcontrol, and long-range mission planning andmonitoring. We also describe the JPL aerobot testbedand the avionics architecture being developed fortesting and validation of aerobot autonomy capabilities.
INTRODUCTION
NASA’s 2003 Solar System Exploration Roadmapstates that aerial platforms could play a key role in theexploration of Mars, Venus and Titan [NASA 2003]. Italso defines advanced autonomy technologies as a highpriority development area for the operation of aerialexploration vehicles. In this paper, we discuss theadvantages of planetary exploration using aerobots, thechallenges involved in aerobot exploration of Titan, and
the required autonomy capabilities. We provide anoverview of the autonomy architecture we aredeveloping, and describe the aerobot testbed and theavionics architecture being developed at JPL for testingand validation of aerobot autonomy capabilities. Wefinalize with a discussion of the current state of thisresearch.
PLANETARY EXPLORATION USINGAEROBOTS
Exploration of the planets and moons of the SolarSystem has so far been done through remote sensingfrom Earth, fly-by probes, orbiters, landers and rovers.Remote sensing systems, probes and orbiters can onlyprovide non-contact, low to medium resolution imageryover a limited number of spectral bands; landersprovide high-resolution imagery and in-situ datacollection and analysis capabilities, but only for a singlesite; while rovers allow imagery collection and in-situscience across their path. The crucial drawback ofground-based systems is their limited coverage: in pastor planned exploration missions, the rover range hasvaried from approximately 130 m (for the 1997Sojourner mission) to 1 km (projected for the 2003Mars Exploration Rovers), to tens of kilometers (for theLunokhod rovers).
While the data collected through these variousapproaches has been invaluable, there is a strategic gapin current exploration technologies for systems that cancombine extensive coverage with high-resolution datacollection and in-situ science capabilities. For planetsand moons with an atmosphere, this gap can be filledby aerial vehicles. In the Solar System, in addition toEarth, the planets Venus and Mars, the gas giants(Jupiter, Saturn, Uranus and Neptune) and the Saturnmoon Titan have significant atmospheres. Aerialvehicles that have been considered for planetaryexploration include airplanes and gliders, helicopters,balloons [Kerzhanovich 2002] and airships. Flight timefor gliders depends heavily on wind and updraftpatterns, which in turn constrain their surface coverage,
AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Tech17 - 19 November 2003, Denver, Colorado
2American Institute of Aeronautics and Astronautics
while airplanes and helicopters expend significantenergy resources simply staying airborne [Elfes 2001,Elfes 2003]. These considerations point towards the useof lighter-than-atmosphere (LTA) systems for planetaryexploration due to their potential for extended missionduration, long traverses, and extensive surface coveragecapabilities. To date, only two LTA vehicles have everflown outside of Earth, as part of the Soviet VEGAmission to Venus in 1985 [VEGA 1985]. Two identicalballoons, each 3.4 m in diameter with a 6.7 kg gondola,flew at an altitude of 53 – 55 km, where theatmospheric temperature and pressure are Earth-like (10– 30oC, 0.5 atm). The balloons had no propulsionsystem, so they simply traveled with the winds. Theenvelopes were constructed from Teflon material towithstand the sulfuric acid in the atmosphere. BothVEGA balloons were delivered inside an entry vehicleand executed aerial deployment and inflation afterarrival, while descending under a parachute. Pressure,temperature and wind speed data were returned for twodays until the onboard batteries ran out of power. It isnot known how much longer the balloons flew afterthat.
Robotic airships have unique capabilities that makethem ideal candidates for airborne planetaryexploration. Airships have modest power requirements,and combine the long-term airborne capability ofballoons with the maneuverability of airplanes orhelicopters. Their controllability allows precise flightpath execution for surveying purposes, long-range aswell as close-up ground observations, station-keepingfor long-term monitoring of high-value science sites,transportation and deployment of scientific instrumentsand in-situ laboratory facilities across vast distances tokey science sites, and opportunistic flight pathreplanning in response to the detection of relevantsensor signatures. Furthermore, robotic airships providethe ability to conduct extensive surveys over both solidterrain and liquid-covered areas, and to reconnoiter sitesthat are inaccessible to ground vehicles.Implementation of these capabilities requires achievinga high degree of vehicle autonomy across a broadspectrum of operational scenarios.
Interest in unmanned airships, primarily for 1)advertising [Foster 2003], 2) military surveillance andintelligence gathering [Boschma 1993], and 3) high-altitude communication platforms [Rehmet 2000], hasbeen growing in the last decade. Small remotely-pilotedairships are becoming commercially available,primarily for advertising [Foster 2003]. Autonomousrobotic airships, however, have only very recentlystarted to be developed. The leading projects in thisarea are AURORA [Elfes 2001], which focuses onautonomy technologies for terrestrial unmanned
airships for environmental research and monitoring;SAA LITE [Boschma 1993], which has developedhighly capable teleoperated surveillance platforms;LOTTE [Kröplin 2002], which addresses new designsand materials for long-term mission solar-poweredunmanned airships; and the JPL aerobot researchdescribed in this paper. Generally speaking, it can besaid that current systems use GPS-based motion controlfor accurate flight trajectory following, while othercapabilities, particularly those related to visualnavigation or long-term mission planning andexecution, are still in their infancy. An integrated set ofenabling technologies for autonomous aerobotnavigation and aerial exploration is currently notavailable, and is the core focus of the research beingdeveloped by the authors.
Figure 1: Near-infrared images of the surface of Titan,taken with the Hubble Space Telescope. Bright areasindicate the possible presence of a continental mass,while dark areas indicate possible oceans of liquidethane and methane. Source: NASA/University ofArizona.
THE CHALLENGE OF TITAN
NASA’s 2003 Solar System Exploration Roadmap alsoidentifies a follow-on Titan mission with an in-situvehicle as a high priority after the Cassini-Huygensmission. Titan is the largest moon of Saturn, with a
3American Institute of Aeronautics and Astronautics
radius of 2,575 km. It has a thick atmosphere with asurface density of 5.55 kg/m3 (4.6 times the density ofthe Earth's atmosphere at sea level), with a nominalcomposition of 95% nitrogen, 3% methane and 2%argon. The surface pressure is approximately 1.5 bar,and the gravity at the surface is 1.35 m/s2
(approximately 1/7 of the gravity of Earth). The surfacetemperature is approximately –180o C.
Although the lower atmosphere is expected to be clearand highly transparent, the upper atmosphere has athick haze that shrouds the surface of Titan from visualobservation. Consequently, very little is known ofTitan’s geography and geology. Recent Hubble SpaceTelescope (HST) observations (Fig. 1) in the near-infrared spectrum (0.85 to 1.05 microns) indicate thepossible existence of a continent composed of solidrock and frozen water ice, and of liquid bodiespotentially composed of liquid ethane and methane[Hall 2002a, Hall 2002b, Lorenz 2000]. Additionallong-term observations have also provided indicationsof weather on Titan, including clouds and storms.
Detailed knowledge of Titan is expected to increasedramatically with the arrival of the Cassini spacecraft atSaturn in July 2004 and the subsequent delivery in 2005of the Huygens probe into the Titan atmosphere.However, Huygens will only investigate Titan at onelocation for a few hours while Cassini will be limited toa few dozen relatively brief flybys of Titan; therefore,many scientific questions will remain unanswered,particularly in the areas of weather and seasonalvariability, subsurface morphology, and thecomposition and distribution of surface organic material[Chyba 1999], leading to the requirement for a follow-on mission.
AEROBOT EXPLORATION OF TITAN
Based on the discussion above, we argue thatexploration of Titan can best be accomplished throughan aerobot, a self-propelled buoyant robotic airship thatcan access most of the world over multi-monthtimescales with minimal consumption of limitedonboard electrical power [Hall 2002a, Hall 2002b] (Fig.2).
The main challenges for exploration of Titan by anaerobot are:
• Large communication latencies, with a round triplight time of approximately 2.6 hours. Thisprecludes both vehicle teleoperation and close,human-in-the-loop, supervisory control.
• Extended communication blackout periods with aduration of up to 9 Earth days, caused by the
rotation of Titan and its orbital occlusion bySaturn.
• Extended duration of the exploration mission,currently projected to be on the order of multiplemonths to a year.
• Operation in substantially unknown environments,with largely unknown wind patterns,meteorological conditions, and surface topography.
Figure 2: An artist’s impression of an aerobot loweringa scientific payload to the surface of Titan. Source: JPL.
These challenges impose the following capabilityrequirements on a Titan aerobot:
• Vehicle safing: the aerobot will have tocontinuously ensure its safety and integrity over thefull duration of the mission and during extendedcommunication blackout, with loose or no humansupervision, and under substantially unknownoperational conditions.
• Autonomous flight: the aerobot will have to executecomplex aerial maneuvers ( includingdeployment/lift-off, landing, hovering/station-keeping, surface approach, and long traverses) withaccuracy and safety, while receiving humanguidance only at the level of goals and pre-computed flight patterns.
• Mapping and self-localization: the aerobot willhave to estimate its motion, localize itself within aregional or global reference frame, and performspatial mapping of its environment without thesupport of a global positioning system andprobably of a magnetic field on Titan. This will beaccomplished through the fusion of inertialmeasurements, vision-based motion estimation,and additional information provided potentially byan infrared Sun tracker, a radio-based Earth trackerand/or a possible future Saturn tracker.
4American Institute of Aeronautics and Astronautics
• Advanced perception: the aerobot has to responddynamically to sensor input, allowing it to detectand avoid atmospheric and topographic hazards,and also to identify, home in, and keep station overpre-defined science targets or terrain featuresthrough visual servoing.
It should be noted that this is an ambitious set ofcapabilities, and that feasible and valuable Titanmissions can be done with less autonomy. Elsewhere,we have discussed a range of mission scenarios andidentified the associated autonomy requirements [Hall2002b].
In the sequence of this paper, we provide an overviewof the design of the autonomy architecture beingdeveloped to address the requirements of Titanexploration through an aerobot. It should be noted thatthis is a multi-year research effort, and that in theremainder of the discussion we will concentrate onthose components of the architecture already developedor currently under development. This includes thedescription of the aerobot testbed being developed atJPL and its avionics architecture.
The work discussed here complements a parallelresearch effort at JPL focused on development of thethermo-mechanical platform required for aerobotoperation in the cryogenic environment at Titan [Hall2003].
AEROBOT AUTONOMY ARCHITECTURE
To address the aerobot autonomy capabilities outlinedabove, we are developing an aerobot autonomyarchitecture that integrates accurate and robust vehicleand flight trajectory control, perception-based stateestimation, hazard detection and avoidance, vehiclehealth monitoring and reflexive safing actions, vision-based localization and mapping, and long-rangemission planning and monitoring. The majorcomponents of this architecture are shown in Fig. 3.
Lower level functions in the autonomy architectureinclude sensor and actuator control, vehicle stateestimation, flight mode control, supervisory flightcontrol, and flight profile execution. Intermediate levelfunctions include vehicle health monitoring, failuredetection and recovery, flight trajectory and profileplanning, and vision based navigation. The latterprovides GPS-independent localization, local andregional mapping, and hazard detection and avoidance(HDA) capabilities. Higher level functions includemission planning, resource management, and missionexecution and monitoring.
Failure Detection& Recovery
Vehicle HealthAssessment
Mission Execution& Monitoring
Earth CommMission Planning
NavigationSensors
ScienceSensors
Flight ProfilePlanning
Flight ProfileExecution
Wind FieldEstimation
Vehicle StateEstimation
ResourceManagement
Aerobot
ActuatorControl
SensorControl
Flight ModeSupervisory Control
Ascend DescendTraverseHoverTakeoff Land
InternalSensors
3D Mapping
HDA
Note: only a representative set of pathways is shown
Figure 3: Aerobot autonomy architecture showing themajor subsystems to be developed.
Our initial research thrust has been concentrated on theactuator and vehicle flight control modules. For that, weare developing:
ß A robust flight control system based on vehicleaerodynamic modeling, system simulation forrobust control law development and testing, andvehicle system identification.
ß Accurate vehicle multi-sensor state estimationmethods, incorporating both inertial and vision-based motion and position estimation.
Once stable and robust flight control has been achieved,our research will focus on vision-based local and globalmapping, target tracking and visual servoingtechniques, internal fault detection and recovery,external hazard detection and avoidance, and flight andmission planning and monitoring.
Aerobot Flight Control
The aerobot flight control system being developed isbased on: (1) system modeling, which includesaerodynamic, airship sensor and actuator, andenvironmental modeling); (2) system identification foraerodynamic parameter estimation; (3) model andcontrol system validation in a physically-basedsimulation environment; and (4) flight testing on theaerobot testbed.
A physically accurate airship aerodynamic model issignificantly different from fixed-wing or rotary-wingaircraft aerodynamic models. Airship dynamic modelshave more in common with submarine hydrodynamics,as the virtual mass and inertia properties of thedisplaced atmospheric volume are substantial whencompared with those associated with the vehicle itself.Additionally, an aerobot is characterized by having
5American Institute of Aeronautics and Astronautics
different flight modes (take-off/landing, station-keeping/hovering, ascent/descent, high-speed cruise,low-speed flight) that require alternative actuatorcontrol strategies and flight control algorithms.Important flight control challenges include non-minimum phase behavior and oscillatory modes at lowspeeds, time-varying behavior due to altitude variations,and variable efficiency of the actuators depending onaerobot speed [Gomes 1990, Elfes 2001].
Wind disturbances will be dealt with using a robustcontroller design. Pose (position and orientation in 6DOF) and motion estimation will initially be done byfusion of IMU and GPS data, allowing assessment ofthe vehicle flight control and trajectory followingaccuracy. To move towards GPS-independentoperation, an image-based vehicle motion and positionestimation system will be developed, and a multi-sensorstate estimation filter will be used to fuse inertial andvisual navigation estimates.
Aerodynamic Model Development
The aerobot aerodynamic model developed has thefollowing features: all bodies are considered rigid,ascent/descent and trajectory studies have beenaddressed, attitude dynamics are included, the planet ismodeled as a flat plane, and thrusting devices aremodeled by (potentially articulated) point force/torquevectors.
The fundamental aerodynamic equation is given by:
M v’ = Fd (v) + A (v) + G + P
where M is the inertia matrix, v is the velocity vectorwith respect to the center of volume, F d are thecentrifugal and Coriolis forces, A are the aerodynamicforces due to the hull and control surfaces, G are thegravitational and buoyancy forces, and P are thepropulsion forces and moments.
The plant model incorporates initial estimates ofaerodynamic coefficients and added (virtual) masscoefficients from theory and tests reported in theliterature [Gomes 1990]). These will be refined throughsystem identification procedures. The complete modelhas been implemented in Simulink (Fig. 4), and initialactuator control studies have been performed. In thesimulation models, lateral and vertical components ofwind gust profiles can be applied. Additionally, sensormodels and estimators (IRU, accelerometers, gyro),actuator models (first order lag model), and trackingcontrol laws (proportional, derivative, and feedforward)are included. Further details, as well as a discussion of
the specific aspects of the kinematic and dynamicmodels developed, are provided in [Quadrelli 2003].
Kinematic &DynamicModels
Control andActuatorModels
GuidanceSystem
Sensors Modelsand State
Estimators
EnvironmentalModels
Figure 4: Aerobot system model implemented inSimulink. The subsystems modeled are: (1) Guidance,(2) Control and Actuator Models, (3) EnvironmentalModels (including wind disturbances and atmosphericcharacteristics), (4) Kinematic and Dynamic Models,(5) Sensor Models and State Estimators.
Aerobot Simulation
Development and testing of the aerobot models andflight controllers will increasingly be supported by ahigh fidelity, physically accurate simulationenvironment. This simulation environment provides alow-cost approach for rapid testing and prototyping ofcontrol, guidance and navigation algorithms beforeflight evaluation on the aerobot. Additionally, it allowsprojection of algorithm and system performance fromEarth to the Titan environment. With the modelsvalidated under Earth flight conditions, the simulationsystem provides the capability to predict theperformance of a Titan aerobot in the Titan atmosphere(Fig. 5).
Vehicle Simulation
Path/AttitudeControl Models
Area/Global Guidance & Navigation Models
High-Fidelity Aerobot Simulation
Earth AirshipTest Platform
PerformancePrediction
Validation Data
Titan Models
Earth Models
TITAN PERFORMANCE
Vehicle Simulation
Path/AttitudeControl Models
Area/Global Guidance & Navigation Models
High-Fidelity Aerobot Simulation
Earth AirshipTest Platform
PerformancePrediction
Validation Data
Titan Models
Earth Models
TITAN PERFORMANCE
Figure 5. The simulation framework provides foraerodynamic model and flight controller validationbefore subsequent flight testing and evaluation on theJPl aerobot testbed; it also allows prediction of systemperformance at Titan.
The aerobot simulation environment will be based onthe high-fidelity spacecraft simulation framework of theDarts/Dshell tool [Biesiadecki 1997]. This includes the
6American Institute of Aeronautics and Astronautics
Darts (Dynamics Algorithms for Real-TimeSimulation) real-time, flexible-body, multibodydynamics package and the Dshell (Darts shell) tool forintegrating reusable hardware and environmentalmodels with Darts to develop high-fidelity spacecraftengineering simulations. This system is the basis for theROAMS (Rover Analysis Modeling & Simulation)planetary rover simulator and the DSENDS (DynamicsSimulator for Entry Descent and Surface landing) entry,descent and landing simulator [Balaram 2002]. Anextensive set of models developed for theseapplications is available for immediate use by the Titanaerobot simulation effort.
Vision-Based Navigation
The aerobot will rely heavily on vision for stateestimation, hazard detection and avoidance, navigation,and mapping. To progress towards GPS-independentoperation over extended travel distances, an image-based vehicle motion and position estimation systemwill be developed using a multi-sensor state estimationapproach. For this, a multi-sensor Kalman filter will beused to fuse inertial navigation measurements(rotational velocities and linear accelerations) from theIMU (inertial measurement unit) with surface relativemotion estimates derived from image-based motionestimation (IBME). IBME makes use of feature imagelocations from feature tracking and the aerobot statefrom the Kalman filter to make these estimates[Roumeliotis 2002].
THE AEROBOT TESTBED
The prototype aerobot testbed being developed at JPL isbased on an Airspeed Airship AS-800B (Fig. 6). Theairship specifications are: length of 11 m, diameter of2.5 m, total volume of 34 m3, two 2.3 kW (3 hp) 23 cm3
(1.4 cu inch) fuel engines, double catenary gondolasuspension, control surfaces in an “X” configuration,maximum speed of 13 m/s (25 kts), maximum ceilingof 500 m, average mission endurance of 60 minutes,static lift payload of 10 kg asl, and dynamic lift payloadof up to 16 kg asl. The avionics and communicationsystems are installed in the gondola.
The aerobot avionics is built around the PC-104+computer architecture (Fig. 7). The PC-104+ stack iscomposed of a CPU board running the onboard avionicssoftware, a serial board interface to the navigationsensors and pan/tilt unit, a timer/counter board forreading pulse width modulated (PWM) signals from ahuman safety pilot and generating PWM signals basedupon control surface commands from the avionicssoftware, and an IEEE 1394 board for sending
commands to, and reading image data from, thenavigation and science cameras.
Figure 6: The JPL aerobot testbed has a length of 11 m,a diameter of 2.5 m, and a static lift payload of 10 kg.
PC-104+
CPU Stack
ControlSurfaces
Inte
grat
ion
Boa
rd
OVERRIDE
Pilot Control72MHz
Serial Rx/Tx900MHz
Li+
Batteries
ScienceCamera
Video Tx426MHz
Video Tx434MHz
Payload Cameras
Cut-AwayRx
NavigationCameraIMU
Compass
LaserAltimeter
GPS
BarometricAltimeter
UltrasonicAnemometer
Nav Avionics
RS232
6-PWM
12V, 6.0W
RS232
5V, 1.0W
RS232
12V, 10W
RS232
12V, 0.3W
2-RS232
5V, 3.2W
RS232
12V, 3.5W
11VDC9Ah
Pan/TiltUnit
RS232
12V, 7.5W
6-PWM
RS232
Figure 7: Avionics architecture.
The navigation sensors consist of an IMU (angularrates, linear accelerations), a compass/inclinometer(yaw, roll and pitch angles), laser altimeter (surfacerelative altitude), barometric altimeter (absolute altitudeagainst reference point), GPS (absolute 3D position)and ultrasonic anemometer (3D wind speed). The visionsensors consist of two down-looking navigationcameras (to gather imagery used for motion andposition estimates that feed into the navigationsoftware) and a science camera (mounted on the pan/tiltunit that acquires imagery used for science-basedprocessing). The science camera is also used foraerobot navigation, but at a higher functional level, i.e.,the navigation camera processing provides inputs intothe Kalman filter used for vehicle state estimation while
7American Institute of Aeronautics and Astronautics
the science camera processing provides inputs intovehicle functions such as go-to or hover over a site ofinterest.
Wireless serial modems provide a two-way linkbetween the aerobot and a ground station. High levelcommands from a human user are sent to the aerobotfrom the ground station and system telemetry is sentfrom the aerobot to the ground station for display andarchival purposes. Two wireless video transmitters onthe aerobot provide a one-way link from the aerobot tovideo receivers at the ground station to send imageryfrom the navigation and science cameras to the groundfor display on video monitors and data archival. Thehuman safety pilot can send vehicle control commandsvia a one-way 72 MHz wireless link using a hand-heldtransmitter to onboard receiver. There is a controlsignal switching box onboard the aerobot that allowsthe routing of either human or robot commands to besent to the vehicle control actuators on an actuator byactuator basis. In addition, the safety pilot can alwaysreassert “pilot override” control over the aerobot. As afinal safety layer, a gas release valve on the aerobotenvelope can be activated by a standalone one-waywireless link from the ground to the aerobot, causingloss of lift and a forced but graceful landing.
Figure 8: Screenshot of the ground station showingaerobot state estimate telemetry from the onboardavionics system.
The ground station (Fig. 8) is composed of a laptop, thewireless data and video links, video monitors andVCRs. The final component of the ground station is adifferential GPS (DGPS) base station that providesdifferential corrections to the GPS receiver onboard theaerobot, allowing vehicle 3D position estimates with anaccuracy on the order of centimeters. This is
particularly important in the early stages of flightcontrol system development and in later stages forvalidation and accuracy assessment of the vision-basednavigation system.
Figure 9: Streaming video imagery from the down-looking navigation camera.
FLIGHT TESTING
We have recently initiated flight testing of the aerobottestbed. An operational subset of the avionics systemhas been demonstrated in tethered outdoor tests (Figs. 8and 9). Fig. 10 shows the maiden flight of the JPLaerobot, conducted on September 10, 2003. Test flightsof the aerobot are being conducted at the El Mirage drylake site in the Mojave desert. Currently, the system isteleoperated, but as the complete avionics system istransitioned to the vehicle and the control systems arevalidated in simulation, vehicle flight control will beprogressively switched to the onboard autonomysystem.
CONCLUSIONS
In this paper, we have argued that robotic LTAvehicles, or aerobots, have a strategic potential for the
8American Institute of Aeronautics and Astronautics
exploration of planets and moons with an atmosphere,such as Venus, Mars and Titan. We argued thataerobots can provide geographically extensive sciencedata at high resolution and over varied terrains, to adegree that cannot be matched by surface-bound roversor other aerial vehicles. We discussed the challenges foroperation of an aerobot under Titan conditions, and thecorresponding autonomy requirements. We outlined anarchitecture for a substantially autonomous aerobot, anddiscussed the aerobot vehicle and flight control systemsbeing developed. We also described the airship testbedand the avionics architecture being developed at JPL forvalidation of aerobot autonomy capabilities. Our nextsteps include transitioning the aerobot control fromteleoperation to autonomous flight trajectory execution,and incorporation of vision-based navigationcapabilities.
ACKNOWLEDGMENTS
The authors would like to acknowledge the help andsupport of: Marco Quadrelli, who developed theaerobot kinematic and dynamic models; J. BobBalaram, who is leading the development of the aerobotsimulation environment; Eric A. Kulczycki, who wasinstrumental in aerobot propulsion and flight testing;and Lee Magnone and Michael S. Garrett, for supportin avionics development. The research described in thispaper was performed at the Jet Propulsion Laboratory,California Institute of Technology, under a contractwith the National Aeronautics and SpaceAdministration (NASA), and administered through theIntelligent Systems (IS) Program. The views andconclusions contained in this document are those of theauthors and should not be interpreted as representingthe official policies, either expressed or implied, of thesponsoring organizations.
Figure 10: Test flight of the JPL aerobot underteleoperated control, conducted at the El Mirage drylake in the Mojave desert. The photos show liftoff,ascent, and flight of the aerobot.
9American Institute of Aeronautics and Astronautics
REFERENCES
[Balaram 2002] J. Balaram et al. "DSENDS - A High-Fidelity Dynamics and Spacecraft Simulator forEntry, Descent and Surface Landing", IEEE 2002Aerospace Conf., Big Sky, Montana, March 2002.
[Biesiadecki 1997] J. Biesiadecki, A. Jain, M. L. James.“Advanced Simulation Environment forAutonomous Spacecraft”, in In ternat ionalSymposium on Artificial Intelligence, Robotics andAutomation in Space (i-SAIRAS'97), Tokyo,Japan, July 1997.
[Boschma 1993] J. H. Boschma. “The DevelopmentProgress of the U.S. Army’s SAA LITE UnmannedAirship”. In Proceedings of the AIAA Lighter-than-Air Systems Technology Conference, AIAA,September 1993.
[Chyba 1999] C. Chyba et al, “Report of the Pre-bioticChemistry in the Solar System Campaign ScienceWorking Group”, Technical Report, JPL, 1999.
[Elfes 2001] A. Elfes et al. “Perception, Modelling andControl for an Autonomous Robotic Airship”. InH. I. Christensen, G. D. Hager (eds.), Sensor BasedIntelligent Robots, Springer-Verlag, Berlin, 2001.
[Elfes 2003] A. Elfes et al. “Robotic Airships forExploration of Planetary Bodies with anAtmosphere: Autonomy Challenges”. Journal ofAutonomous Robots, v. 14, n. 2/3. KluwerAcademic Publishers, The Netherlands, 2003.
[Foster 2003] N. Foster. Airspeed Airships.http://www.airspeedairships.com, 2003.
[Gomes 1990] S. B. V. Gomes. “An Investigation of theFlight Dynamics of Airships with Application tothe YEZ-2A. PhD thesis, College of Aeronautics,Cranfield University, 1990.
[Hall 2002a] J. L. Hall, V. V. Kerzhanovich, J. A.Jones, J. A. Cutts, A. A. Yavrouian, A. Colozza, R.D. Lorenz. “Titan Airship Explorer”, inProceedings of the 2002 IEEE AerospaceConference, IEEE, Big Sky, MT, March 2002.
[Hall 2002b] J. L. Hall, A. Elfes, T. Spilker, V.Kerzhanovich, J. F. Montgomery. “Titan Aerobots:An Overview of Mission Scenarios and RequiredAutonomy Technologies”. Whitepaper, JPL 2002.
[Hall 2003] J. L. Hall. “Titan Aerobot Design”. JPLinternal document, 2003.
[Kerzhanovich 2002] V. V. Kerzhanovich, J. A. Cutts,H. W. Cooper, J. L. Hall et al. “Breakthrough inMars Balloon Technology”, in Proceedings of theWorld Space Congress / 34th Scientific Assemblyof the Committee on Space Research (COSPAR),Houston, TX, USA, October, 2002.
[Kröplin 2002] B. Kröplin. “Solar Airship LOTTE”.Technical Report, Institute for Statics andDynamics of Aerospace Structures, University ofStuttgart, Germany, 2002.
[Lorenz 2000] R. D. Lorenz. “Post-Cassini Explorationof Titan: Science Rationale and MissionConcepts”. In Journal of the British InterplanetarySociety, vol. 53, UK, 2000.
[NASA 2003] Solar System Exploration.http://spacescience.nasa.gov/admin/divisions/se/SSE_Roadmap.pdf.
[Quadrelli 2003] M. Quadrelli. “Titan AerobotSimulation Model”, JPL, August 4, 2003.
[Rehmet 2000] M. A. Rehmet et al. “RecentDevelopments on High Altitude Platforms”. InProceedings of the 3rd International AirshipConvention and Exhibition (IACE 2000), TheAirship Association, UK. Friedrichshafen,Germany, July 2000.
[Roumeliotis 2002] S. I. Roumeliotis, A. E. Johnson,and J. F. Montgomery. “Augmenting InertialNavigation with Image-Based Motion Estimation”.In Proceedings of the 2002 IEEE InternationalConference on Robotics and Automation,Washington, DC, 2002.
