Project AURORA: Development of an Autonomous Unmanned Remote Monitoring Robotic Airship

Project AURORA:

Development of an Autonomous Unmanned Remote

Monitoring Robotic Airship

Alberto Elfes, Samuel S. Bueno, Marcel Bergerman

and Josué Jr. G. Ramos

Robotics and Computer Vision Laboratory,

Automation Institute, Informatics Technology Center

Caixa Postal 6162 -Campinas, SP Brazil 13083-970

email: [e-mail]

Sérgio Bittencourt Varella Gomes

LTA Brasil Ltda.

Rio de Janeiro, RJ Brazil

Abstract There exists an immense potential for the utilization of robotic airships as low speed, low altitude aerial vehicles in exploration, monitoring, and transportation tasks. This article discusses Project AURORA - Autonomous Unmanned Remote mOnitoring Robotic Airship which focuses on the development of the control, navigation, sensing, and inference technologies required for substantially autonomous robotic airships. Our target application areas include the use of robotic airships for environmental, biodiversity, and climate research and monitoring. Based on typical mission requirements, we present arguments that favor airships over airplanes and helicopters as the ideal platforms for such missions. We outline the overall system architecture of the AURORA robotic airship, discuss its main subsystems, and mention the research and development issues involved.

1 Introduction

Recent advances in the areas of sensors, sensorial interpretation, and control and navigation systems have fostered interest in the development of intelligent semi-autonomous robotic systems. While unmanned semi-autonomous ground and underwater robotic vehicles are being increasingly used in a variety of applications, much less progress has been made in robotic aerial vehicles.

Although unmanned aerial vehicles (UAVs) play an important role in military reconnaissance and surveillance missions [8], and agencies such as NASA are employing aerial vehicles as platforms for environmental and climate research [9], these systems are typically flown using remote control during all critical phases of the mission, while using onboard navigation systems for flight path execution. Relatively little work has been done so far towards the development of robotic aerial vehicles that could plan and execute extended missions with a significant degree of autonomy.

Unmanned aerial vehicles have an enormous application potential, which includes traffic monitoring, urban planning, inspection of man-made structures and archaeological site prospection. They can also be utilized in environmental, biodiversity, and climate research and monitoring (henceforth denoted simply by environmental monitoring), including sensoring and monitoring of forests, national parks, and ecological sites, land use survey, agricultural studies, livestock inventory, limnological studies, identification of atmospheric and limnological pollutants and their sources, characterization of plants and animals and long-term variability studies.

Today, many of these applications are pursued based on remote sensing data obtained from balloon-borne sensors, satellite imagery, or manned aircraft-based aerophotogrammetric studies. These sources of information have significant drawbacks, however. Balloons are not maneuverable and, therefore, the sensed area cannot be directly controlled. Satellite imagery available for civilian applications is limited in terms of its spatial (pixel) resolution, of the spectral bands available, of its temporal coverage, and of the geographical area (spatial swath) covered. Manned aerophotogrammetric or aerial inspection surveys, while giving the mission planner control over most of the variables mentioned above, are very costly in terms of the aircraft to be used, crew time, maintenance time, etc.

We believe that the development of unmanned, semi-autonomous robotic aerial vehicles will ultimately allow the acquisition of aerial-gathered information in such a manner as to give the user the ability to choose the spatial and time resolution of the data to be acquired, to define the appropriate geographical coverage, and to select the sensorial systems of relevance for a specific data-gathering mission, while doing so at more readily affordable costs. This will lead directly to an expansion of scientific and civilian uses of aerial data and to significant social and economic benefits deriving from this expansion.

Among the aircraft used as UAVs, by far the most commonly employed are reduced-scale fixed-wing vehicles (airplanes), followed by rotary-wing (helicopter) aircraft. Airships, also known as lighter-than-air (LTA) vehicles, are only recently becoming a focus of research [1] and finding applications in several niche areas [7].

In the work discussed in this article, our driving interest is in robotic aerial vehicles for environmental, biodiversity, and climate research and monitoring. As we will discuss later, airships outrank other kinds of aircraft for missions with this kind of profile. It is in this context that Project AURORA has been started. Our goal is to develop a robotic airship with significant levels of autonomy during all phases of its operation. This includes the ability to perform mission, navigation, and sensor deployment planning and execution, failure diagnosis and recovery, and adaptive replanning of mission tasks based on real-time evaluation of sensor information and constraints on the airship and its surroundings.

Unmanned airships available today are remotely controlled, usually with small payload capabilities. The few exceptions with larger payload are the MAMBOW 3 [13], and the SASS-LITE systems [1], this one presenting some onboard navigation capabilities. Project AURORA does not focus primarily on advancing the underlying LTA technologies as such, but rather on the modeling, control, sensing, inference, and navigation capabilities required by LTA vehicles. It provides, therefore, an important contribution to the field of semi-autonomous robotic airships.

This article is divided as follows: in Section 2 we describe typical mission characteristics, and why they lead to airships as the ideal aerial vehicles for environmental monitoring. Section 3 provides a description of the physical principles of airship operation, and of their propulsion systems and actuators. In Section 4 we describe the project subsystems. In Sections 5 and 6 we describe the hardware and software architectures and the human-machine interface. In Section 7 we outline airship modeling and control. In Sections 8 and 9 we discuss autonomous and vision-based navigation. In Section 10 we present our preliminary conclusions.

2 Why an Airship?

Most of the environmental monitoring studies mentioned in the Introduction have profiles that require low speed, low-altitude data gathering platforms. Such vehicles must also be able to hover above an area and must have extended airborne capabilities, for long duration studies; generate very low noise and turbulence, so as not to disturb the environment that is being measured and monitored; generate very low vibration, so as to reduce sensor noise and hardware malfunction; be able to take off and land vertically, so that maintenance and refueling can be done without the need for runways, and also so that remote, difficult to access regions with limited logistics support can be monitored; be highly maneuverable; have a large payload to weight ratio; and have a low operation cost. Of the four possible aerial vehicles - airplanes, helicopters, airships, and balloons - the last ones are not considered here because they are not maneuverable. Table 1 compares the three remaining vehicles with respect to the requirements set above, where high compliance with each requirement is indicated by three marks, and low or no compliance by one mark. One can infer from Table 1 that airships are, on the average, more suited to environmental monitoring tasks than airplanes or helicopters. This is mainly due to the fact that the largest part of their lift comes from aerostatic forces, rather than aerodynamic ones. Therefore, an airship does not need to spend a lot of energy to float in the air, only to move around between locations or to counteract the drift caused by the wind. The consequence is that airships need smaller engines than airplanes and helicopters for propulsion, which in turn produce less noise, vibration, and turbulence, and consume less fuel.

Table 1 -
Comparison of aerial vehicles as platforms for environmental
research and monitoring missions.
(high =

Project requirement

Airplane

Helicopter

Airship

Low operation cost

Long endurance

Hovering capability

Payload to weight ratio

High maneuverability

Low noise and turbulence

Vertical take-off and landing

Low fuel consumption

Low vibration

3 Physical Principles of Airship Operation

In this section we describe the physical principles of airship operation. The description is very general, and valid for both manned and unmanned vehicles. Naturally, simplifications apply for unmanned systems.

3.1 Aerostatic Lift

The airship has as its main source of lift what is called aerostatic lift, i.e., lift that is independent of flight speed. Aerostatic lift is calculated by multiplying the volume of air displaced by the lifting gas, by the difference in density between such gas and air. That is also why such force is still known today as buoyancy. Therefore, only gases that are lighter than air can be used, helium being the most commonly used today.

3.2 Ballonets

The discussion above is valid for a vehicle restricted to a certain height. As it ascends through the atmosphere, air gets less dense due to the decrease in the ambient pressure. In order therefore to retain the same buoyancy force upwards, the lifting gas needs to expand to proportionately higher volumes. This is achieved by letting air out of the ballonets, which are simply bags of air inside the main (lifting gas) envelope (Figure 1).

At this stage it is important to note that all current airships are of the so-called non-rigid or pressure airship type. They keep their streamlined shape through pressurization, the internal gas pressure being only 0.5 to 1% higher than that of the surrounding air ambient pressure.

When air is let out of the ballonets in climbs or is pumped into them in descents, the associated amounts are those just adequate to keep the prescribed pressure differential range mentioned above. Once an altitude is reached at which there is no more air to be expelled from the ballonets, then further ascent would only be possible by valving off the lifting gas, an expensive and usually forbidden operation in the case of helium. It is therefore said that the vehicle has attained its pressure height, i.e., its operational ceiling.

Figure 1:
Ballonet operation in a traditional non-rigid airship.

3.3 Temperature Effects

Temperature affects LTA vehicles in two different ways. Firstly, aerostatic lift, like any quantity directly proportional to gas density, decreases as the temperature increases (and vice-versa) if everything else stays the same. As a consequence, an airship designed to have a certain pressure height under given temperature and pressure conditions may find it is short on performance if suddenly required to operate from hot and high airfields. Secondly, in a way unique to LTA, the lifting gas can acquire a temperature significantly higher (or lower) than that of its surrounding air. Called superheat, this is a phenomenon which can occur, for example, after long exposure to direct sunlight and/or a rapid ascent through the atmosphere. The resulting lifting gas expansion usually generates a relatively fast increase in the buoyancy force. This in turn can sometimes become quite a nuisance for the steady control of the vehicle.

3.4 Heaviness and Lightness

When the airship is powered by fuel-burning engines, it will undergo a mass decrease over a given period of operation. If not properly accounted for, it can come to land with the buoyancy force far exceeding the weight of the complete vehicle. In that case, the landing operation can become quite a hazardous activity, to be avoided at all costs, since it could ultimately mean valving off expensive helium gas. To prevent such occurrence, three means are usually resorted to. Firstly, the most common of all which is to take off with the vehicle carrying enough ballast. This results in the vehicle weight being higher than the buoyancy force on landing, even after the burning of the trip’s fuel. Secondly, using vectored thrust, i.e., deflecting the thrust propellers downwards so as to help bring the airship down when such mechanical facility is available. And thirdly, partly replacing the mass of fuel burned along the way by increasing the amount of water ballast through the use of heat exchangers placed in the exhaust gas stream out of the propulsion engines.

3.5 Propulsion and Actuation

Operation of an airship starts, from a theoretical point of view, with what is available to its forerunner, the balloon: release of gas, affecting the buoyancy force, and release or addition of ballast, affecting the total mass of the vehicle. In practice, helium is not usually released for the reasons previously discussed and the following features are used:

• aerodynamic control surfaces

• vectored thrust

• bow and/or stern thrusters

The simple conclusion from the previous points is that the airship uses the same aerodynamic controls as conventional aircraft plus its power units for the same purpose. In fact, like any buoyant vehicle, below a certain speed, aerodynamic controls do not work at all and control reversal, i.e. a non-minimum phase behavior, is quite a reality. At very low speeds, going up or down is then achieved by vectored thrust and going left or right demands the use of differential thrust in the right and left hand side propellers, or the use of stern thrusters when available.

4 Project AURORA

AURORA builds on previous work done by the authors on autonomous underwater vehicles [4] and remotely piloted helicopters [11]. It is conceived as a three-phase project (see Table 2), with a sequence of prototypes to be developed capable of successively higher mission times and mission range, while also displaying increasing levels of autonomy, evolving from mainly teleoperated to mainly autonomous systems. Naturally, successful completion of the three phases does not mean that AURORA III will be capable of full commercial operation; nonetheless, the airship-based environmental monitoring concept will have been fully demonstrated and will serve as the basis for future development.

Table 2 -
Project phases.

	AURORA I	AURORA II	AURORA III
Typical mission duration (h)	1-2	8	> 24
Typical distance (km)	1-10	10-50	> 100
Typical payload	10	50	> 100
capacity (kg)

The major physical subsystems of AURORA I are: the airship; the onboard control and navigation subsystems, including the internal sensors, hardware, and software; the communications subsystem; the mission sensors; and a mobile base station (Figure 2). By internal sensors we understand those atmospheric, inertial, positioning, and imaging sensors required by the vehicle to accomplish its autonomous navigation tasks. Mission sensors are those selected for specific aerial data-gathering needs, and are not discussed in details. The other subsystems are described in the sequence.

Figure 2:
AURORA I project components.

AURORA I is conceived as a proof-of-concept system, to be used in low-demanding applications. The LTA platform is the AS800 by Airspeed Airships [14] (Figure 3), a non-rigid, 9 m long, 3 m diameter, 30 m³ airship equipped with two vectorable engines on the sides of the gondola and four control surfaces at the stern, arranged in ‘X’. Its useful payload capacity is around 10 kg at sea level.

The onboard control and navigation subsystems are responsible for sensor data acquisition and actuator control, based on flight profiles uploaded to it from the ground. The airship hardware consists of an onboard computer, microprocessors, internal sensors, and actuators. The software consists of a 3-layer architecture.

Figure 3:
The AURORA I airship: Airspeed Airship's AS800.

The communications subsystem is composed of radio links which transmit data and commands between the airship and the base station. This system includes also video links for the transmission of imagery captured by cameras mounted onboard.

The mobile base station is composed of a processing and communication infrastructure, a portable mooring mast, and a ground vehicle for equipment transportation.

5 Hardware

The internal sensor suite used for flight path execution purposes includes both inertial navigation sensors - compass, accelerometers, inclinometers, and gyroscopes - and a GPS (global positioning system) receiver. Cameras mounted on the airship’s gondola have the purpose of providing aerial images to the operator on the ground, and serving for visual navigation based on geographical features of the terrain.

Figure 4 presents the hardware architecture integrating the airship’s onboard equipment; the remote control unit, which can be used to override the commands sent to the actuators, is not shown. A compass, inclinometer, and GPS receiver are directly connected, via serial ports, to a PC 104 computer. All other control, navigation, and diagnosis sensors (engine speed, altitude, control surface position, wind speed, accelerometers, fuel and battery level, and engine temperature) and actuators (engines and control surfaces) are connected to a microprocessor. The microprocessor is responsible for sending all sensor readings to the computer, and for sending commands to the actuators.

Figure 4:
Onboard and ground hardware configuration.

6 Software Architecture and Human-Machine Interface

The software architecture consists of a 3-layer structure, combined with a high-level data flow programming method and system development environment.

As airships have relatively large time constants, at the lowest level we use a soft real-time operating system based on a reduced Linux kernel.

The complexity of the robotic system being developed requires a deliberative-reactive intermediate-level process control and communications architecture, where different subsystems can run independently and as separate threads, while able to exchange information and activate or inhibit each other. For that we are using the Task Control Architecture [12], which gives us a convenient structure to handle multiple subsystems running in a distributed computing environment.

As the higher-level, overarching control structure we are considering a layered, multi-rate approach similar to the ATLAS architecture [4], which allows nested control cycles to be run at progressively lower speeds, but at increasing levels of competency.

The human-machine interface (HMI) provides the communication and visualization mechanism between the operator and the navigation system on board the airship. In AURORA I the operator uses this interface to receive imagery, as well as other flight data acquired onboard, and to define the flight profile, given as a set of locations and altitudes. The HMI will be progressively enhanced as the project advances, and will include integration with a geographical information system.

Figure 5:
Virtual reality airship simulator.

As part of the HMI we have developed a physical model-based virtual reality airship simulator [10] (Figure 5). It allows for experimentation over the Internet with airship control during all phases of the flight, including take-off and landing. The simulator is based on a very accurate dynamic model of the airship, outlined in the sequence, and incorporates real-world topographical information of selected regions. We are utilizing the simulator to validate control strategies and navigation methods, for pilot training, and for mission design and pre-evaluation.

7 Dynamic Modeling and Control System

As part of the Project AURORA we have surveyed dynamic modeling for non-rigid airships, to be used for model-based control and navigation system development. To establish a workable mathematical model of airship flight a number of considerations have to be taken into account, as it differs from the usual conventional aircraft models:

• the LTA vehicle displaces a very large volume and its virtual (added) mass and inertia properties become significant, i.e., it behaves as if it had a mass and moments of inertia substantially higher than those indicated by conventional physical methods;

• the airship’s total mass can change considerably in a very short time. In a climb or descent maneuver for example, this is due to ballonet deflation or inflation respectively;

• in order to reasonably accommodate the constantly changing Center of Gravity (CG) position, a characteristic which is unique to airship flight, the airship motion has to be referenced to a system of orthogonal body axes fixed in the vehicle with the origin at the Center of Volume (CV). The CV is also assumed to coincide with the gross Center of Buoyancy (CB).

The dynamic model was derived originally for a buoyant ROV Later, it was modified to reproduce the peculiarities of airships, such as the Westinghouse’s YEZ-2A [6]. In building the full non-linear 6DOF mathematical model only two limiting assumptions were made at the outset for practical reasons:

• the airship forms a rigid body such that aeroelastic effects can be ignored;

• the airframe is symmetric about the longitudinal plane such that both the CV and the CG lie in the plane of symmetry.

For the type of airship we are using, the dynamic model is given by [5, 6]

(1)

where:

• n is the 6x1 velocity vector, that contains the three linear velocities

  u,

  v,

  w and three angular velocities

  p,

  q,

  r, all written with respect to the body-fixed reference frame. For navigation purposes, they must be transformed to an Earth fixed (North-East-Up) inertial reference frame.

• M

• F

  _d

• A

• G

• P

Based on this model, we have analyzed the airship’s motion and found four interesting control challenges: non-minimum phase behavior and oscillatory modes at low speeds, time-varying behavior due to altitude variations and fuel burning, and the variable efficiency of the actuators depending on the airship’s speed. More specifically, regarding the last one, at very low speeds aerodynamical forces do not contribute to the airship maneuverability, i.e., the control surfaces have negligible influence on the airship motion. In this case, adjusting the propellers’ thrust and vectoring angle is the main type of actuation. On the other hand, at moderate to high speeds, it is the aerodynamical forces who contribute most to the motion, with propeller angle having little effect.

The AURORA I control system is designed as a 3-layer hierarchical structure. At the bottom level, actuators provide the means for maneuvering the airship along its course. At the intermediate level, control algorithms with different gains are available; they command the actuators based on decisions taken on the top level. The top level, a shared supervisory one, decides on which control algorithm is to be activated, its set-point, and the related actuators, depending on which is the current flight part - take-off, cruise, turning, landing, hovering, etc.

We are investigating overall system robustness at two different levels: at the intermediate level, ensuring that each control algorithm is robust to modeling errors and external disturbances, such as winds; and at the top level, ensuring that the control algorithms and related actuators are robustly selected, and ensuring bumpless transfer between them.

The control and navigation systems depend crucially on accurate information regarding the airship’s location in six DOFs with respect to an Earth-fixed reference frame. A continuously updated estimate of the airship position and pose is obtained through a state estimator, implemented as a Kalman filter, having as inputs the sensor data collected from the inertial sensors and the GPS receiver, and as output the continuous optimal estimate of the airship location in a least squares sense.

Other control issues being dealt with include shared control between the operator and the airship, so that the operator can indicate broad directions to follow while the control system makes sure the local trajectory is followed correctly, or even take over system control in case of unforeseen events; bumpless control transfer between the operator and the control system; and approaches to deal with unforeseen problems, such as communication degradation or loss.

8 Autonomous Navigation

On top of the control system aforementioned, another level is added, to cope with autonomous navigation. We are currently developing a navigation system that is able to automatically compute flight plans based on mission profiles, while obeying the airship's dynamic constraints and incorporating limited information about atmospheric conditions. We will subsequently expand this work to incorporate weather information in the flight planning process, to avoid unstable weather regions.

To provide global mission planning and monitoring at the highest logical level, we are investigating the use of a probabilistic planner. The planner will crbeate the appropriate action sequences for successful mission execution, and will oversee activation of the deliberative and reactive components of the underlying software architecture. The planning interface will allow the operator to define appropriate mission goals and parameters.

For system diagnosis and failure recovery, we are developing a Bayesian causal network modeling approach, based on our previous work [4]. This allows the system to use probabilistic reasoning to infer potential failure causes from sensorial data, and to subsequently execute recovery or graceful degradation actions.

(b)

(c)

(d)

Figure 6: (a) Aerial image taken in the infra-red spectrum; (b) Computing a reference trajectory to be followed by the aerial vehicle along the river center; (c) Identification of two potential pollution plume candidates; (d) Adjustment in the height of the aerial vehicle along the reference trajectory to allow for closer observation of plume candidates.

9 Perception and Vision-Based Navigation

As we mentioned earlier, most UAV systems today operate through a combination of remote control and programmed execution of fixed mission and flight plans. Extending the capabilities of a robotic aircraft so that it is able to dynamically respond to its state estimate and to the information obtained from its internal and mission sensors, and thereby closing the control cycle and allowing the robot to adjust its mission and flight plans accordingly, is one of the key challenges towards broader UAV deployment.

Sensor-based adaptive navigation of a robotic aircraft requires several perceptual competencies, including visual position registration and egomotion estimation, landmark identification and triangulation-based motion control, visual following of large-scale structures such as roads and rivers, and sensor-based target identification and tracking, among others.

Many of these competencies require an active approach to perception control, and this has led us to explore the application of the Dynamic Perception framework [2, 3] to airborne visual navigation problems. For that, we are using the Inference Grid model, a tessellated, multi-property random field approach to the representation of spatially distributed information [2], to encode sensor inferences and also incorporate digital terrain map data. Imagery obtained from the aircraft sensors is interpreted using probabilistic sensor models, and Bayesian and Maximum Likelihood estimators are used to update the cell state estimates of the random field model. Information from the airship position estimators is used to help register the data gathered against the world models being built. To actively control both the execution of a specific task and the acquisition of the information required for that task, we are applying the dual control work discussed in [3]. For that, the Dynamic Perception framework allows us to define Loci of Interest (LI) in the Inference Grid space, and Loci of Observation (LO) in the sensor parameter and airship location space, as ways to represent spatial regions about which data is required and regions in the sensor control space that define how the required data will be observed.

To provide a brief illustration of adaptive vision-based guidance of an aerial vehicle, consider the infra-red image shown in Figure a. The goal of the system is to follow the river and search for potential pollution plumes for closer scrutiny. By decomposing the infra-red color image in its HSV components to autonomously identify the river in the image and subsequently applying an image distance transform, a reference trajectory along the center of the river is computed (Figure 6b). By following the river trajectory and using now the intensity component of the image, as well as texture extraction operators, the system is able to identify two potential plume candidates - which correspond to Loci of Interest (Figure 6c). The need to observe these regions more closely and to diminish the uncertainty in the classification of the plumes - expressed using an entropy metric - leads to the computation of the appropriate Loci of Observation and consequently to an adjustment of the height of the aerial vehicle along the reference trajectory (Figure 6d).

10 Conclusions

This article is a preliminary report on Project AURORA, describing some of our initial work towards the development of a semi-autonomous robotic airship. Unmanned robotic airships present an enormous untapped potential for applications in low-speed, low-altitude exploration, surveillance, and monitoring, as well as telecommunication relay platforms. The control and navigation systems for semi-autonomous airship operation being developed in Project AURORA are a step towards the fulfillment of such unexplored potential. The AURORA airship will serve as an environmental monitoring platform, with tasks including surveying and monitoring of forests and national parks, land use survey, air composition and pollution measurements above cities and industrial sites, and biodiversity and limnological studies.

Acknowledgments: This project is partially supported by the Brazilian funding agencies FAPESP and CNPq.

• [1] J.H. Boschma. "The development progress of the U.S. Army’s SAA LITE, Unmanned Robot Airship." Proceedings of the AIAA Ligther-Than-Air Systems Technology Conference, Sep. 1993.
• [2] A. Elfes. "Dynamic control of robot perception using stochastic spatial models." In Information Processing in Mobile Robots, Schmidt (ed.), Springer-Verlag, 1991.
• [3] A. Elfes. Robot navigation: integrating perception, environmental constraints and task execution within a probabilistic framework. In Reasoning with Uncertainty, Doorst, van Lambalgen, Voorbraak (eds.), Springer-Verlag, 1996.
• [4] A. Elfes. Autonomous Underwater Vehicle (AUV) Architecture IBM T. J. Watson Research Center Internal Report, 1990.
• [5] S. B. V. Gomes and J.J.G. Ramos. Airship dynamic modeling for autonomous operation. In Proc.IEEE Int. Conf. on Robotics and Automation, Leuven, Belgium, May 1998.
• [6] S.B.V. Gomes. An investigation of the flight dynamics of airships with application to the YEZ-2A. Ph.D. Thesis, Cranfield Institute of Technology, Oct. 1990.
• [7] E. Mowforth. An introduction to the airship Airship Association Publication no. 3, The Airship Association Ltd., 1991.
• [8] K. Munson, ed. Jane's unmanned aerial vehicles and targets Jane’s Information Group Limited, Surrey, 1996.
• [9] National Air and Space Administration. Theseus Project http://www.hq.nasa.gov/office/mtpe/ Theseus.html, Oct. 1997.
• [10] J.J.G. Ramos et al. "Project AURORA: autonomous unmanned remote monitoring robotic airship." 2nd Int. Airship Convention and Exhibition, Bedford, England, Jun. 1988.
• [11] J.J.G. Ramos and O. Neves. Environment for unmanned helicopter control system development: application examples. In Proc.IEEE 4th Conference on Control Applications, Albany, NY, Sep. 1995.
• [12] R. Simmons. Structured control for autonomous robots. IEEE Trans. on Robotics and Automation, 10 (1): 34-43, 1994.
• [13] K. Tairo. MAMBOW 3: Unmanned outdoor blimp for multi remote sensing. SKYPIA Co. Ltd., Tokyo, Japan. Private communication.
• [14] N. Wells. Practical operation of remotely piloted airships. In Proc. 11th AIAA Lighter-than-Air Systems Technology Conference, Clearwater Beach, FL, May 1995, paper no. 95-1634.

Publication Dates

• Publication in this collection
  08 Oct 1998
• Date of issue
  Apr 1998