Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Automation of Mobility and Navigation 16.4 Navigation Control and Interaction with the Environment 285

tivariable nonlinear system with strong coupling in

some control loops. Autonomous helicopter control has

been a classical control benchmark and many model-

based control techniques have been applied, including

multi-PID controllers, robust control, predictive con-

trol, and nonlinear control. The signiﬁcance of each of

these methods for practical implementations is still an

open question. Safe landing on mobile platforms and

transportation of loads by means of several UAVs over-

coming the payload limitation of individual UAVs are

open challenges. Very recently the joint transportation

of a load by several helicopters has been demonstrated

in the AWARE (acronym of the project platform for

autonomous self-deploying and operation of wireless

sensor-actuator networks cooperating with aerial ob-

jects) project [16.28].

There are also approaches in which learning from

skilled human pilots or teleoperators plays the most sig-

niﬁcant role. In these approaches, fuzzy logic [16.29],

neural networks [16.30], neurofuzzy techniques, and

other artiﬁcial intelligence (AI) techniques are ap-

plied. Control theory and AI techniques have also

been combined. Thus, for example, Takagi–Sugeno

fuzzy systems have been applied to learn from hu-

man drivers by generating closed-loop control systems

that can be analyzed and tuned by means of stabil-

ity theory. These methods have been applied to drive

autonomously trucks (Fig. 16.9) and heavy machines

at high speed [16.31] by estimating the position of

the vehicle by means of the fusion of GPS and dead-

reckoning with simple vehicle models. The system has

Fig. 16.9 Autonomous 16 t Scania truck

been applied to test the tires of vehicles navigating au-

tonomously in testing tracks.

The general challenges are the analysis and design

of reliable controltechniques that couldbe implemented

in real time at high frequency in the onboardprocessors,

providing reactivity to perturbations while maintaining

acceptable performance. This is particularly challeng-

ing when considering small or very small vehicles, such

as micro-UAVs with important limitations in onboard

processing. The application of micro-electromechanical

systems (MEMS) to implement these control systems

is an emerging technology trend. The practical appli-

cation in real time of fault-detection techniques and

fault-tolerant control systems to improve reliability is

another emerging trend.

16.4 Navigation Control and Interaction with the Environment

The consideration of interactions with the environment

is also an important problem in mobility and navigation

automation. These interactions can also be represented

by means of loops closed at different frequencies, as

showninFig.16.10. The vehicle control described in

the above paragraphs is also embedded in this ﬁg-

ure. Reactivity dominates the higher-frequency loops,

which also require higher bandwidth in communication

channels (inner loops, towards the right in the ﬁgure),

while deliberation is the main component of the lower-

frequency loops, which typically have lower bandwidth

requirements in communication channels (outer loops,

towards the left in the ﬁgure). The inner loops can be

considered as the lower lever in the control hierarchy,

while the outer loops are the higher levels in this hier-

archy. This chapter will not provide details on mobile

robot control architectures, but will merely describe the

main interactions.

Environment perception is based on the use of

sensors such as cameras, radars, lasers, ultrasonic,

and other range sensors. Thus, cameras and radars

have been applied extensively for the guidance of au-

tonomous cars. The processing of the images leads

to the computation of relevant environment features

that can be used to guide the vehicle by means of

visual servoing techniques (image-based visual servo-

ing). Alternatively, the features can be used to compute

the position/orientation of the vehicle and then apply

position-based visual servoing. The stability of the vi-

sual control loop in the guidance of vehicles has been

studied by several authors (see, for example, [16.32]).

The main drawbacks of these methods are robustness to

Part B 16.4

286 Part B Automation Theory and Scientiﬁc Foundations

Task

generation

Mission Complexity Simplicity, analysis

ReactivityDeliberation

Increasing bandwidth

Increasing frequency

Cognitive

models

Path

generation

Geometric

models

Trajectory

generation

Distance

position

Vehicle

control

Features

extraction

Fig. 16.10 Vehicle control and decision loops. The width of the arrows indicates the frequency of the loops and re-

quired bandwidth in communication. The inner loops (right in the ﬁgure) correspond to higher-frequency loops with

higher bandwidth requirements, while the outer loops (left in the ﬁgure) correspond to lower-frequency loops with lower

bandwidth requirements

illumination changes and real-time constraints. Laser-

based environment perception techniques are also used

in navigation [16.33], as shown by the 2005 DARPA

Grand Challenge (DGC) [16.34]. However, even laser

measurements have some drawbacks in outdoor envi-

ronments; for example, dust clouds could be treated as

transient obstacles or weeds and large rocks cannot be

differentiated.

Compensation for suchdifferentenvironmentalcon-

ditions plays an important role. The use of several

sensors and the application of sensor data-fusion meth-

ods signiﬁcantly improve robustness against changes in

these conditions. Thus, for example, most autonomous

vehicles in the DGC (see Fig.16.11 for the 2007 Urban

Grand Challenge)applied sensordata-fusion techniques

for autonomous navigation. In some cases environment

perception can substitute or complement GPS position-

Fig. 16.11 Winners of the Urban Grand Challenge, Novem-

ber 2007 [16.35]

ing, overcoming problems related to the visibility of

satellites and degradation of the GPS signal.

The above-mentionedtechniques to computethe po-

sition of the vehicle with respect to the environment

can also be applied to generate trajectories in these

environments. Trajectory generation methods can even-

tually consider the kinematics and dynamic constraints

of the vehicles in order to obtain trajectories that can

be realistically executed with the vehicle control sys-

tem described in the previous section. The relevance

of these techniques depends on the characteristics of

the vehicles. Thus, for example, they are very rele-

vant in the navigation of ground wheeled vehicles with

conventional car-like locomotion systems or in ﬁxed-

wing airplanes, but could be ignored in omnidirectional

ground vehicles navigating at low speeds.

The computation of distances and positions of

the vehicle with respect to the environment can also

be used to obtain geometric models of the environ-

ment. Particularly mapping techniques can be applied.

Moreover, many probabilistic simultaneous localization

and mapping techniques (SLAM)havebeenpro-

posed and successfully applied in robotics in the last

decade [16.36]. Most implementations have been car-

ried out by using lasers in two-dimensional (2-D)

environments. However, the methods for the application

of SLAM in 3-D environments are also promising.

The results of vehicle position and environment

mapping can be used for the planning of vehicle mo-

tion. The planning problem consists of the computation

of a path for the vehicle from a starting conﬁguration

(position/orientation) to a goal position/orientation con-

ﬁguration, avoiding obstacles and minimizing a cost

Part B 16.4

Automation of Mobility and Navigation 16.4 Navigation Control and Interaction with the Environment 287

index usually related to the length of the path [16.37].

Today, many different techniques for automatic path

planning can be applied. In the basic problem a geomet-

ric model of the vehicle is assumed and the model of the

environment is assumed to be completely known (map

known) and static, without other vehicles and mov-

ing obstacles. Many path planning methods are solved

in the conﬁguration space C, deﬁned by the conﬁg-

uration variables q ∈

that completely specify the

position and orientation of the vehicle. Then, given

start

, q

goal

∈C, the problem is to ﬁnd a sequence of

conﬁgurations in the obstacle-free conﬁguration space

∈C

free

connecting q

start

and q

goal

The problem can be solved by searching in the

discretized space. Well-known methods are visibility

graphs and Voronoi diagrams. In these methods the

connectivity of the free space is represented by means

of a network of one-dimensional (1-D) curves. These

methods have been extensively applied in 2-D en-

vironments. The consideration of 3-D models adds

signiﬁcant complexity for execution in real time. An-

other well-known strategy consists of searching in an

adjacency graph of the free-space cells, obtained by dis-

cretization of the environment model into occupancy

cells. These methods need the implementation of graph

searching algorithms, such as A*, to ﬁnd a solution

and they are usually quite time consuming. The appli-

cation of multiresolution techniques greatly improves

the computational efﬁciency of these methods. On the

other hand, in recent years, randomized methods and

particularly so-called rapidly exploring random trees

(RRT) [16.38],have been used to explore the free space,

obtaining good results. This method makes it possible

to explore high-dimensional conﬁguration spaces, and

even to include different constraints. Several of these

techniques have been applied and extensively tested in

the DGC 2005 competition.

There are also methods based on the optimization

of potential functions attracting the vehicle to the goal

(global minimum) by considering at the same time the

effect of repulsive forces exerted by the obstacles. The

obvious difﬁculty of these latter methods, initially pro-

posed for real-time local collision avoidance [16.39],

is the existence of local minima. The potential-based

methods have also been applied for motion planning,

eventually combined with other planning methods such

as the space-cell decomposition mentioned above. Dif-

ferent potential functions have been proposed to avoid

the local minima problem in these methods [16.40].

The extensions of the basic motion planning prob-

lem described above include the consideration of the

nonholonomic and dynamic constraints of the vehi-

cle [16.41]. Other extensions include the uncertainties

in the models of the vehicle and the environment, as

well as the motion of other vehicles and obstacles in the

environment.

The stability of reactive navigation is studied

in [16.42], where Lyapunov techniques, input/output

stability (conicity criterion), and frequency response

methods are applied to study the stability of the nav-

igation of an autonomous ground vehicle by using

ultrasonic sensors. The stability is related to the param-

eters of the reactive navigation such as the sensor range

and the velocity. The inﬂuence of the time delay, due

to communication and computation, on the stability of

the reactive navigation is also considered. The analy-

sis is based on the deﬁnition of the perception function

p =ψ(d,θ), where d and θ are, respectively, the dis-

tance and angle at which an obstacle is detected. The

values of p are provided to the closed-loop controller.

Then, the feedback controller u =ϕ(p)isappliedtothe

vehicle with model given by (16.1).

Planning under uncertaintyhas been a researchtopic

for many years. In the classical planning methods the

world is assumed to be deterministic, and the state ob-

servable. The uncertainty can be taken into account

by considering stochastic Markov decision processes

(MDPs) with observable states, which leads to stochas-

Fig. 16.12 The Aurora robot spraying a greenhouse

Part B 16.4

288 Part B Automation Theory and Scientiﬁc Foundations

tic accurate models. If the state is assumed to be only

partially observable, then partially observable Markov

decision processes (POMDPs) can be used to consider

stochastic inaccurate models.

The outer loop (the upper level in the control archi-

tecture) in Fig.16.10 is based on the consideration of

cognitive models of the environment obtained from the

geometric representations, knowledge extracted from

the sensorial information, i.e., identiﬁcation of partic-

ular objects in the environment (i.e., trafﬁc signals),

and previous human knowledge onthe existing relations

between these objects and the vehicle navigation.

A key architectural issue is the appropriated com-

bination of reactivity and planning. This is related to

the interaction between the different levels or control

loops in Fig.16.10 and has a signiﬁcant impact on the

methods to be applied. Thus, some control architec-

tures are based on the application of very simple motion

strategies, without considering uncertainties or mobile

objects in the environment, and providing reactivity

based on real-time sensing of the environment. Reac-

tive techniques can be implemented in behavior-based

control architectures to navigate in natural environ-

ments without models or with a minimal model of the

environment. Figure 16.12 shows the Aurora mobile

robot [16.43] which uses a behavior-based architec-

ture to navigate in greenhouses usingultrasonic sensors.

Other architectures are based on dynamic planning

by incorporating environment information in real time

and producing a new plan that reacts appropriately

to the new information. In the architecture presented

in [16.42] planning techniques based on the kinemat-

ics model of the vehicle are used to generate parking

maneuvers for articulated vehicles (Fig.16.8).

One of the main general trends is the integration

of control and perception components into embedded

systems that can be networked using wired or wireless

technologies, leading to cooperating objects, with sens-

ing and/or actuation capabilities based on sensor fusion

methods that allow full interaction with the environ-

ment. Open challenges are related to the development

of tools for the analysis and design of these systems.

16.5 Human Interaction

In practice mobility automation requires the interven-

tion of humans at a certain level. The key point is the

level of interaction. Thus, with regards to Fig.16.10,

the human can be provided with information from the

cognitive models, which encode expertise from other

operators, and use this information to decompose a mis-

sion into tasks to be planned by the task generator

module in the Fig.16.10.

However, if the task planner does not exist, the hu-

man operator can interact with the second outer loop.

In this case the human operator can use the map of the

environment displayed on a suitable display to generate

a sequence of waypoints for the vehicle using an appro-

priate interface to specify these waypoints by means of

a joystick or a simple mouse. The human operator can

also generate in the next inner loop a suitable trajectory

assisted by computer tools to visualize in an appropri-

ate way the distances to the surrounding obstacles and

check the suitability of these trajectories to be executed

by the particular vehicle being commanded. Going to

the right in the control loops of Fig. 16.10, the operator

could directly provide commands to the vehicle control

loop by observing signiﬁcant environment features in

the images provided by the onboard camera or cameras.

The above-mentioned interactions involve hardware

and software technologies to provide appropriate senso-

rialfeedback(visual,audio...)tothehumanpilotsand

generate actions at different levels from direct guidance

to waypoint and task speciﬁcation.

At this point it is necessary to distinguish between

human interventions onboard the vehicle and operation

in a remote teleoperation station or from suitable re-

mote devices such as personal digital assistants (PDAs)

or even mobile phones involving the communication

system, as shown in Fig. 16.13. The ﬁrst approach

can be considered as a compromise solution between

vehicle full driving automation, which removes the

driver from the control loop, to assisted driving to im-

prove efﬁciency and reduce accidents, as mentioned in

the Introduction. Integration of automatic functions in

conventional cars has been a trend in the last years.

Autonomous parking of conventional vehicles is an ex-

ample of this trend. On the other hand, the development

of mixed autonomous/manual driving cars seems a suit-

able approachfor thegradual integration ofautonomous

vehicles on regular roads. Furthermore, the automation

of functions in aircraft navigation is also well known.

In the following, the remote teleoperation of vehi-

cles is considered. Figure 16.13 illustrates teleoperation

schemes.

The low-level classical teleoperation approach con-

sists of the presentation to the remote teleoperator of

Part B 16.5

Automation of Mobility and Navigation 16.5 Human Interaction 289

images from a camera or cameras mounted on the front

of the vehicle. Then, the vehicle is guided manually us-

ing joysticks, pedals or similar interfaces to the ones

existing in the driving position onboard the vehicles.

This approach has many problems; for example, the

images may be degraded due to bandwidth instability,

leading to poor spatial resolution and variable update

rates, degrading the perception of the motion. Further-

more, an important problem is the presence of delays in

both the images and sensor data sent to the operator and

the operator commands sent back to the vehicle. These

delays may generate instability ofthe teleoperation con-

trol loop.

Various technologies have been proposed to over-

come these problems. Thus, instead of showing in the

display the data and images from the vehicles, it is pos-

sible to process these data to extract relevant features to

be displayed. Note that this approach can be included

in the inner (faster) control loop of Fig. 16.10. The

computation of distances in the second loop could be

relevant too. Moreover, the displaying of geometric rep-

resentations of the environment around the vehicle (see

also Fig.16.13) combined with the images (augmented-

reality technologies) could also be very useful.

A classical method to reduce the harmful effect

of delays in the transmission of images is the use of

predictive displays with a synthetic graphic of the ve-

hicle obtained by means of a simulation model. The

graphic is overlaid on the real delayed images from the

vehicle. This representation leads the operator to per-

Actuation

generation

Actuation

generation

Transmission

Information

acquisition

TransmissionPresentation

Planning

reacting

control

Perception

active perception

Fig. 16.13 Human interaction loops

ceive in advance the effect of his commands, which can

be used to compensate for delays. This can be easily

combined with the augmented reality mentioned above.

Thus, real-world imagery is embedded within a display

of computer-generated landmarks or objects represent-

ing the same scene. Thecomputer-generatedcomponent

of a display can be updated immediately in response to

control inputs from the human operator, providing rapid

feedback to the operator. If a model of the environment

is known, it can be stored in databases and rendered

based, for example, on the current GPS position of the

vehicle.

Obviouslythe navigation conditions may have a sig-

niﬁcant effect on the operators. Thus, it has been

pointed out that UAV operators may not modify their

visual scanning methods to compensate for the non-

recreated multisensory cues. In order to improve the

perception of the operators, haptic and multimodal in-

terfaces (e.g., tactile and auditory) have been proposed.

Multimodal interfaces may be used not just to com-

pensate for the teleoperator’s sensory environment, but

more generallyto reducecognitive-perceptual workload

levels. Thus, for example, [16.44] have found that audio

and tactile messages can improve many aspects of ﬂight

control and overall awareness of the situation in UAV

teleoperation.

The teleoperation methods presented above greatly

depend on communication between the vehicle and

the teleoperation station. The development of mobile

communication in the last decade has changed the sit-

Part B 16.5

290 Part B Automation Theory and Scientiﬁc Foundations

uation when compared with the existing technologies

when the ﬁrst autonomous vehicles where developed in

the 1980s. Obviously, the communication technologies

to be applied greatly depend of the level of interven-

tion of the human teleoperator in the architecture of

Fig.16.13. Thus, high-bandwidth communication is es-

sential at the lower- and high-frequency loops, where

vehicle onboard autonomy is low and generation of

teleoperation commands depends on observation of the

images and information from other sensors onboard the

vehicle. However, if the vehicle has onboard auton-

omy, the communication with the user does not require

high bandwidth. Thus, for example, GSM (global sys-

tem for mobile communications) and GPRS (general

packet radio service) have been used for communicat-

ing the vehicles with the users through their mobile

phones and PDAs. Furthermore, Wi-Fi (IEEE 802.11)

has been applied for communication with the vehicle

at low velocity and short range. An emerging trend is

the application of ad hoc mobile networks to take into

account the particular mobility of vehicles.

The development of models of the loops in

Fig.16.13, as well as new analysis and design tools

using these models, are important challenges to be

addressed. These models should involve not only the

vehicles, the control devices, and communication chan-

nels, but also suitable models of the human perception

and action mechanisms, which typically require signif-

icant experimentation efforts to cope with the behavior

of operators under different working conditions.

16.6 Multiple Mobile Systems

The automation of multiple vehicles offers many ap-

plication possibilities. The interest in transportation is

obvious. A basic conﬁguration of multiple vehicles con-

sists of a leader followed by vehicles in a single row.

This is usually known as platooning [16.45]. The con-

trol of a platoon can be implemented by means of a local

strategy, i.e., each vehicle is controlled from the unique

data received from the vehicle at the front [16.46]. This

approach relies mainly on the single-vehicle control

problem considered above. The main drawback is that

the regulation errors introduced by sensors noises grow

from the ﬁrst vehicle to the last one, leading to os-

cillations. Intervehicle communication can be used to

overcome this problem [16.47]. Then, the distance, ve-

locity, and acceleration with respect to the preceding

vehicle are transmitted in order to predict the position

and improve the controller by guaranteeing the stabil-

ity of tight platoon applications [16.48]. Intervehicle

communication can also be used to implement global

control strategies.

The formation of multiple vehicles is also useful

for applications such as searching and surveying, ex-

ploration and mapping, hazardous material handling

systems, active reconﬁgurable sensing systems, and

space-based interferometry. The advantages when com-

paring with single-vehicle solutions are increased efﬁ-

ciency, performance, reconﬁgurability, and robustness.

An added advantage is that new formation members

can be introduced to expand or upgrade the formation,

or to replace a failed member. Thus several applica-

tions of aerial, marine, and ground vehicle formations

have been proposed. In these formations, the members

of the group of vehicles should keep user-deﬁned dis-

tances from other group members. The control problem

consists of maintaining these user-deﬁned distances.

Formation control involves the design of distributed

control laws with limited and disruptedcommunication,

uncertainty, and imperfect or partial measurements. The

most common approach is the leader-follower [16.49].

This approach has limitations when considering the re-

liability of the leaders and the lack of explicit feedback

from the follower to the leader. Then, if the follower

is perturbed by some disturbances, the formation can-

not bemaintained. Thereare also alternativeapproaches

based on virtual leaders [16.50], which is a reference

point that moves according to the mission. The stability

of the formation has been studied by many researchers

that have proposed robust controllers to provide insen-

sitivity to possibly large uncertainties in the motion

of nearby agents, transmission delays in the feedback

path, and the effect of quantized information. There are

also behavior-based methods [16.51], which are often

inspired by biology, where formation behaviors such

as ﬂocking and following are common. Different be-

haviors are deﬁned as control laws for reaching and/or

maintaining a particular goal. An emerging trend in for-

mation control is the integration of obstacle avoidance

into control schemes.

Other approaches are based on the consideration of

teams of robots describing different trajectories to ac-

complish tasks. Furthermore, having a team with multi-

ple heterogeneous vehicles offers additional advantages

due to the possibility of exploiting the complementar-

ities of vehicles with different mobility attributes and

Part B 16.6

Automation of Mobility and Navigation 16.6 Multiple Mobile Systems 291

also different sensors with different perception func-

tionalities. The vehicles need to be coordinated in time

(synchronization) to accomplish missions such as mon-

itoring. Spatial coordination is required to ensure that

each vehicle will be able to perform its plan safely and

coherently, regarding the plans of the others. Assum-

ing that multiple robots share the same world, a path

should be computed for each one that avoids collisions

with obstacles and with other robots. Some formula-

tions are based on the extension of single-robot path

planning concepts such as the conﬁguration space. If

there are nr robots and each robot has a conﬁguration

space C

, i = 1,...,nr, the state space is deﬁned as

the Cartesian product X = C

×C

×···×C

and the

obstacle region in X is

obs





i=1

obs



∪

⎛

⎝



ij,i=j

obs

⎞

⎠

(16.4)

where X

obs

and X

obs

are the robot–obstacle and the

robot–robot collision states. The problem is to ﬁnd

a continuous path in the free space from the initial state

to the goal state, avoiding the obstacle region deﬁned

by (16.4). The classical planning algorithms for a sin-

gle robot with multiple bodies [16.37] could be applied

without adaptation in case of a centralized planning that

takes into account all robots. The main concern, how-

ever, is that the dimension of the state space grows

linearly in the number of robots. Complete algorithms

require time that is at least exponential in dimension.

Sampling-based algorithmsare morelikely to scale well

in practice when there many robots, but the resulting

dimension might still be too high. There are also decou-

pled path planning approaches such as the prioritized

planning that considers one robot at a time according to

a global priority.

On the other hand, cooperation is deﬁned in the

robotic literature as a joint collaborative behavior that

is directedtowardsome goal inwhichthere is acommon

interest or reward. According to [16.54], given some

task speciﬁed by a designer, a multiple-robot system

displays cooperative behavior if, due to some underly-

ing mechanism, there is an increase in the total utility

of the system. Cooperative perception can be deﬁned

as the task of creating and maintaining a consistent

view of a world containing dynamic objects by a group

of agents, each equipped with one or more sensors.

Cooperative vision perception has become a relevant

topic in the multirobot domain, mainly in structured

environments [16.55, 56]. In [16.57] cooperative per-

ception methods for multi-UAV system are proposed.

Fig. 16.14 Coordinated ﬂights in the COMETS project

[16.52]

These methods have been implemented in the archi-

tecture designed in the COMETS project (acronym of

the project real-time coordination and control of mul-

tiple heterogeneous unmanned aerial vehicles) [16.58]

(Fig.16.14). Cooperative perception requires integra-

tion of the results of individual perception. Each robot

extracts knowledge by applying individual perception

techniques, and the overall cooperative perception is

performed by merging the individual results. This

approach requires knowing the relative position and ori-

entation of the robots. If the GPS signal is not available,

position estimation based on environment perception

should be applied [16.59,60]. The cooperation of mo-

bile entities also involves the generation of appropriated

motion of the involved entities. In [16.61] the coopera-

Fig. 16.15 Experiment of the CROMAT system for

the cooperation of aerial and ground robots [16.53],

http://grvc.us.es/cromat

Part B 16.6

292 Part B Automation Theory and Scientiﬁc Foundations

Multiple vehicle

systems

Formations

Teams

Homogeneous teams,

swarms,

heterogeneous teams

Formation stability,

guidance,

obstacle avoidance

Temporal coordination: synchronization,

spatial coordination: path planning,

task planning

Fig. 16.16 Methods in multiple vehi-

cle systems

tion is categorized into:swarm type, dealingwith a large

number of homogeneous robots, usually involving nu-

merous repetitions of the same activity over a relatively

large area; and intentional cooperation, usually requir-

ing a smaller number of possibly heterogeneous robots

(Fig.16.15) performing several distinct tasks. In these

systems the multirobot task allocation problem [16.62]

is appliedto maximizethe efﬁciencyof theteam anden-

sure proper coordination among team members to allow

them to complete their mission successfully. Recently,

a very popular approach to multirobot task allocation

has been the application of market-based negotiation

rules by means of the contract net protocol [16.63,

64].

Figure 16.16 shows the different types of multiple-

vehiclesystems and the methods that are applied in each

type.

Communication and networking also play an im-

portant role in the implementation of control systems

for multiple unmanned vehicles. The star-shaped net-

work conﬁguration with all the vehicles linked to the

control station with an unshared link only works well

with small teams. When the number of vehicles grows

it could be necessary to apply wireless heterogeneous

networks with radio nodes mounted at ﬁxed ground sta-

tions, on ground vehicles, and in UAVs, and the routing

techniques allow any two nodes to communicate either

directly or through an arbitrary number of other nodes

which act as relays. Furthermore, when there is little

or no infrastructure, networks could be formed in an

ad hoc fashion and information exchanges occur only

via the wireless networking equipment carried by the

individual UAVs.

Finally, it should be noted that the wireless net-

working of teams of robots with sensors and actuators

embedded in the infrastructure is a new research and

development trend with many potential applications.

The AWARE project (http://www.aware-project.net)

is developing a new platform for the cooperation

of autonomous aerial vehicles with ground wireless

sensor–actuator networks. This platform will have

self-deployment and self-conﬁguration features for op-

eration in sites without sensing and communication

infrastructure.

16.7 Conclusions

Mobility and navigation have been very relevant topics

in automation. Thus, automation of mobility plays an

important role in factory automation. The automation

of the transportation of people and goods in noncon-

trolled environments is more difﬁcult and its complexity

depends on the ﬂexibility. This chapter has analyzed

automation of mobility and navigation by focusing on

autonomous vehicles. Then, vehicle motion control has

been examined, and the main problems in navigation

control and interaction of the vehicle with the environ-

ment were also studied. Moreover, taking into account

that practical applications usually require some degree

of human intervention, human interaction and related

technologies were reviewed. The last part of the chapter

was devoted to systems of multiple autonomous vehi-

cles, including formations and ﬂeets of homogeneous

vehicles and also teaming of heterogeneous vehicles.

The control and cooperation of these autonomous vehi-

cles to accomplish tasks is an emerging trend that poses

different challenges.

References

16.1 R. Marín, J. Garrido, J.L. Trillo, J. Sáez, J. Armesto:

An industrial automated warehouse based on

overhead trolleys, MCPL’97 IFAC/IFIP Conf. Manag.

Control Prod. Logist. (Campinas, 1997) pp. 137–142

Part B 16

Automation of Mobility and Navigation References 293

16.2 C.E. Thorpe (Ed.): Vision and Navigation: The

Carnegie Mellon Navlab (Kluwer, Boston 1990)

16.3 M. Parent, A. de La Fortelle: Cybercars: past,

present and future of the technology, Proc. ITS

World Congr. (2005)

16.4 R. Horowitz, P. Varaiya: Control design of an au-

tomated highway system, Proc. IEEE 88(7), 913–925

(2000)

16.5 UAV Forum: http://www.uavforum.com/ (last ac-

cessed March 5, 2009)

16.6 J. Moraleda, A. Ollero, M. Orte: A robotic system for

internal inspection of water pipelines, IEEE Robot.

Autom. Mag. 6(3), 30–41 (1999)

16.7 H.M. Kim, J. Dickerson, B. Kosko: Fuzzy throttle and

brake control for platoons of smart cars, Fuzzy Sets

Syst. 84, 209–234 (1996)

16.8 R.W. Brockett: Asymptotic stability and feedback

stabilization. In: Differential Geometric Control

Theory,ed.byR.S.Millman,R.W.Brockett,

H.H. Sussmann (Birkhauser, Boston 1983)

16.9 C.Y. Chan, H.S. Tan: Feasibility analysis of steering

control as a driver-assistance function in collision

situations, IEEE Trans. Intell. Transp. Syst. 2(1), 1–9

(2001)

16.10 J.H. Hahn, R. Rajamani, L. Alexander: GPS-based

real-time identiﬁcation of tire–road friction co-

efﬁcient, IEEE Trans. Control Syst. Technol. 10(3),

331–343 (2002)

16.11 B. Samadi, R. Kazemi, K.Y. Nikravesh, M. Kabga-

nian: Real-time estimation of vehicle state and

tire-road friction forces, Proc. Am. Control Conf.

(Arlington 2001) pp. 3318–3323

16.12 J. Huang, J. Ahmed, A. Kojic, J.P. Hathout: Control

oriented modeling for enhanced yaw stability and

vehicle steerability, Proc. Am. Control Conf. (Boston

2004) pp. 3405–3410

16.13 A. Kamga, A. Rachid: Speed, steering angle and

path tracking controls for a tricycle robot, Proc.

IEEE Int. Symp. Computer-Aided Control Syst. Des.

(Dearborn 1996) pp. 56–61

16.14 C. deWit, B. Siciliano, G. Bastin: Theory of Robot

Control (Springer, Berlin Heidelberg 1997)

16.15 A. Ollero: Robótica. Manipuladores y Robots

Móviles (Marcombo, Spain 2001), in Spanish

16.16 J. Wit, C.D. Crane, D. Armstrong: Autonomous

ground vehicle path tracking, J. Robot. Syst. 21(8),

439–449 (2004)

16.17 A. Rodríguez-Casta

no, A. Ollero, B.M. Vinagre,

Y.Q. Chen: Setup of a spatial lookahead path track-

ing controller, Proc. 16th IFAC World Congr. (Prague

2005)

16.18 T. Hellstr

om, T. Johansson, O. Ringdahl: Devel-

opment of an autonomous forest machine for

path tracking, Springer Tracts Adv. Robot., Vol. 25

(Springer, Berlin Heidelberg 2006) pp. 603–614

16.19 G. Heredia, A. Ollero: Stability of autonomous ve-

hicle path tracking with pure delays in the control

loop, Adv. Robot. 21

(1), 23–50 (2007)

16.20 DARPA Grand Challenge: Special issue, J. Field

Robot. 23(8/9), 461–835 (2006)

16.21 J.Y. Wang, M. Tomizuka: Robust H∞ lateral con-

trol for heavy-duty vehicles in automated highway

systems, Proc. Am. Control Conf. (San Diego 1999)

pp. 3671–3675

16.22 G.H. Elkaim, M. O’Connor, T. Bell, B. Parkinson:

System identiﬁcation and robust control of farm

vehicles using CDGPS, Proc. ION GPS-97 (Kansas City

1997) pp. 1415–1424

16.23 A. González-Cantos, A. Ollero: Backing-up maneu-

vers of autonomous tractor-trailer vehicles using

the qualitative theory of nonlinear dynamical sys-

tems,Int.J.Robot.Res.28(1), 49–65 (2009)

16.24 A. Astolﬁ, P. Bolzern, A. Locatelli: Path-tracking of

a tractor-trailer vehicle along rectilinear and circu-

lar paths: a Lyapunov-based approach, IEEE Trans.

Robot. Autom. 20(1), 154–160 (2004)

16.25 A. Ollero, L. Merino: Control and perception tech-

niques for aerial robotics, Annu. Rev. Control 28,

167–178 (2004)

16.26 O. Amidi, T. Kanade, K. Fujita: A visual odometer for

autonomous helicopter ﬂight, Robot. Auton. Syst.

28, 185–193 (1999)

16.27 M. Bejar, A. Ollero, F. Cuesta: Modeling and control

of autonomous helicopters. In: Advances in Control

Theory and Application, Lect. Notes Control Inf. Sci.,

Vol. 353, ed. by C. Bonivento, A. Isidori, L. Marconi,

C. Rossi (Springer, Berlin Heidelberg 2007) pp. 1–27

16.28 AWARE Project: http://www.aware-project.net (last

accessed March 5, 2009)

16.29 A. Ollero, A. García-Cerezo, J.L. Martínez,

A. Mandow: Fuzzy tracking methods for mobile

robots. In: Applications of Fuzzy Logic: Towards

High Machine Intelligence Quotient Systems,Vol.9,

ed. by M. Jamshidi, L. Zadeh, A. Titli, S. Boverie

(Prentice Hall, Upper Saddle River 1997) pp. 347–

364, Chap. 17

16.30 G. Buskey, G. Wyeth, J. Roberts: Autonomous he-

licopter hover using an artiﬁcial neural network,

Proc. IEEE Int. Conf. Robot. Autom. (2001) pp. 1635–

1640

16.31 A. Ollero, A. Rodríguez-Castaño, G. Heredia: Anal-

ysis of a GPS-based fuzzy supervised path tracking

system for large unmanned vehicles, Proc. 4th IFAC

Int. Symp. Intell. Compon. Instrum. Control Appl.

(SICICA) (Buenos Aires 2000) pp. 141–146

16.32 F.Conticelli,D.Prattichizzo,F.Guidi,A.Bicchi:

Vision-based dynamic estimation and set-point

stabilization of nonholonomic vehicles, Proc. 2000

IEEE Int. Conf. Robot. Autom. (San Francisco 2000)

pp. 2771–2776

16.33 J. González, A. Stenz, A. Ollero: A mobile robot

iconic position estimator using a radial laser scan-

ner, J. Intell. Robot. Syst. 13, 161–179 (1995)

16.34 M. Buehler, K. Iaguemma, S. Singh: The 2005 DARPA

Grand Challenge, Springer Tracts Adv. Robot.,

Vol. 36 (Springer, Berlin Heidelberg 2007)

Part B 16

294 Part B Automation Theory and Scientiﬁc Foundations

16.35 DARPA Urban Challenge: http://www.darpa.

mil/grandchallenge/images/photos/11_4_07/D2X_

1328.jpg (last accessed March 5, 2009)

16.36 S. Thrun, W. Burgard, D. Fox: Probabilistic Robotics,

Intelligent Robotics and Autonomous Agents (MIT

Press, Cambridge 2005)

16.37 R.C. Latombe: Robot Motion Planning (Kluwer,

Boston 1991)

16.38 S.M. LaValle: Rapidly-exploring random trees:

A new tool for path planning TR 98-11 (Iowa Univ.,

Iowa 1998)

16.39 O. Khatib: Real-time obstacle avoidance for ma-

nipulators and mobile robots, Int. J. Robot. Res.

5(1), 90–98 (1986)

16.40 S.A. Masoud, A.A. Masoud: Motion planning in the

presence of directional and regional avoidance

constraints using nonlinear, anisotropic, harmonic

potential ﬁelds: a physical metaphor, IEEE Trans.

Syst. Man Cybern. Part A, 32(6), 705–723 (2002)

16.41 V.F. Muñoz, A. Ollero, M. Prado, A. Simón: Mo-

bile robot trajectory planning with dynamic and

kinematic constraints, Proc. IEEE Int. Conf. Robot.

Autom., San Diego (1994) pp. 2802–2807

16.42 F.Cuesta,A.Ollero:Intelligent mobile robot

navigation, Springer Tracts Adv. Robot., Vol. 16

(Springer, Berlin Heidelberg 2005)

16.43 A. Mandow, J. Gomez de Gabriel, J.L. Martinez,

V.F. Mu

noz, A. Ollero, A. García-Cerezo: The au-

tonomous mobile robot aurora for greenhouse

operation, IEEE Robot. Autom. Mag. 3(4), 18–28

(1996)

16.44 G.L. Calhoun, M.H. Draper, H.A. Ruff, J.V. Fontejon:

Utility of a tactile display for cueing faults, Proc.

Hum. Factors Ergon. Soc. 46th Annu. Meet. (2002)

pp. 2144–2148

16.45 P. Daviet, M. Parent: Platooning for small public

urban vehicles, 4th Int. Symp. Exp. Robot. (ISER’95)

(Stanford 1995) pp. 345–354

16.46 J. Bom, B. Thuilot, F. Marmoiton, P. Martinet:

Nonlinear control for urban vehicles platooning,

relying upon a unique kinematic GPS, 22nd Int.

Conf. Robot. Autom. (ICRA’05) (Barcelona 2005)

pp. 4149–4154

16.47 Y. Zhang, E.B. Kosmatopoulos, P.A. Ioannou,

C.C. Chien: Autonomous intelligent cruise control

using front and back information for tight vehi-

cle following maneuvers, IEEE Trans. Veh. Technol.

48(1), 319–328 (1999)

16.48 T.S. No, K.-T. Chong, D.-H. Roh: A Lyapunov func-

tion approach to longitudinal control of vehicles in

a platoon, IEEE Trans. Veh. Technol. 50(1), 116–124

(2001)

16.49 J.P. Desai, J.P. Ostrowski, V. Kumar: Modeling and

control of formations of nonholonomic mobile

robots, IEEE Trans. Robot. Autom. 17(6), 905–908

(2001)

16.50 M. Egerstedt, X. Hu, A. Stotsky: Control of mobile

platforms using a virtual vehicle approach, IEEE

Trans. Autom. Control 46, 1777–1782 (2001)

16.51 T. Balch, R.C. Arkin: Behavior-based formation

control for multi-robot teams, IEEE Trans. Robot.

Autom.

14, 926–939 (1998)

16.52 A. Ollero, I. Maza: Multiple Heterogeneous Aerial

Vehicles, Springer Tracts Adv. Robot., Vol. 37

(Springer, Berlin Heidelberg 2007)

16.53 I.Maza,A.Viguria,A.Ollero:Aerialandground

robots networked with the environment, Proc.

Workshop Netw. Robot Syst. IEEE Int. Conf. Robot.

Autom. (2005) pp. 1–10

16.54 Y.U. Cao, A.S. Fukunaga, A. Kahng: Cooperative

mobile robotics: Antecedents and directions, Au-

ton. Robots 4(1), 7–27 (1997)

16.55 T. Schmitt, R. Hanek, M. Beetz, S. Buck, B. Radig:

Cooperative probabilistic state estimation for

vision-based autonomous mobile robots, IEEE

Trans. Robot. Autom. 18(5), 670–684 (2002)

16.56 S. Thrun: A probabilistic online mapping algorithm

for teams of mobile robots, Int. J. Rob. Res. 20(5),

335–363 (2001)

16.57 L. Merino, F. Caballero, J.R. Martínez-de Dios,

J. Ferruz, A. Ollero: A cooperative perception system

for multiple UAVs: application to automatic detec-

tion of forest ﬁres, J. Field Robot. 23(3), 165–184

(2006)

16.58 A. Ollero, S. Lacroix, L. Merino, J. Gancet, J. Wik-

lund, V. Remuss, I.V. Perez, L.G. Gutiérrez,

D.X. Viegas, M.A. González, A. Mallet, R. Alami,

R. Chatila, G. Hommel, F.J. Colmenero, B.C. Arrue,

J. Ferruz, J.R. Martinez-de Dios, F. Caballero: Mul-

tiple eyes in the skies, IEEE Robot. Autom. 12(2),

46–57 (2005)

16.59 K.Konolige,D.Fox,B.Limketkai,J.Ko,B.Stew-

art: Map merging for distributed robot navigation,

IEEE Int. Conf. Intell. Robot. Syst. (2003) pp. 212–

217

16.60 L. Merino, F. Caballero, J. Wiklund, A. Moe,

J.R. Martínez-de Dios, P.-E. Forssen, K. Nordberg,

A. Ollero: Vision-based multi-UAV position estima-

tion, Robot. Autom. Mag. 13(3), 53–62 (2006)

16.61 L.E. Parker: Alliance: An architecture for fault-

tolerant multi-robot cooperation, IEEE Trans.

Robot. Autom. 14(2), 220–240 (1998)

16.62 B.P. Gerkey, M.J. Mataric: A formal analysis and

taxonomy of task allocation in multi-robot sys-

tems,Int.J.Robot.Res.23(9), 939–954 (2004)

16.63 S.C. Botelho, R. Alami: M+: a scheme for multi-

robot cooperation through negotiated task alloca-

tion and achievement, Proc. IEEE Int. Conf. Robot.

Autom. (Detroit 1999)

16.64 B. Gerkey, M. Mataric: Sold: Auction methods for

multi-robot coordination, IEEE Trans. Robot. Au-

tom. 18(5), 758–768 (2002 )

Part B 16