Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Virtual Reality and Automation 15.3 Medical Applications 275

aiming for the ventricles. A distinct popping or punc-

turing sensation is felt as the catheter enters the frontal

horn of the lateral ventricle [15.19].

The performance evaluation from the Immersive-

Touch simulator based on a large sample of neuro-

surgeons was comparable to clinical study ﬁndings, in

which the average distance and standard deviation of

the catheter tip to the foramen of Monroe in both cases

were close [15.18].

15.3.2 Ophthalmic Virtual Automation

Cataract surgery is the most common surgical pro-

cedure performed by ophthalmic residents-in-training

and general ophthalmologists in the USA. With over

1.5million procedures performed every year, it repre-

sents the largest single surgical care provided under

Medicare in the USA.

Cataract surgery involves the removal of the opaci-

ﬁed crystallinelens by phacoemulsiﬁcation andreplace-

ment of the lens with an intraocular lens implant. Over

the years cataract surgery has become a highly tech-

nical procedure requiring a number of very ﬁne skills,

under an operating microscope. It involves the use and

understanding of sophisticated technology, as well as

ﬁne hand–eye coordination movements. The training of

this highly skilled procedure to residents has become

a challenging task for teaching physicians because of

its highly technical nature. Current ways of training res-

idents include:

•

Practising surgical procedures in a wet lab situa-

tion, where surgery is practised on animal eyes.

Though this procedure has some similarity to hu-

man surgery, there are many problems associated

with this technique, which include: (a) Porcine and

bovine models that do not provide a realistic sense

of tissue tension, nor do they mimic the accurate

dimensions of the human eye; (b) Any assessment

of skills acquired during such animal laboratory

procedures with anobserving physician isoften sub-

jective and repetition is difﬁcult.

•

Another technique that has been used extracts eyes

from a postmortem human eye bank to perform pha-

coemulsiﬁcation. These tissues are very difﬁcult to

obtain, and are fairly expensive to procure.

•

A common technique, the apprenticeship tech-

nique, is a see one, do one, teach one approach.

However, apprenticeship is not the best approach,

especially for a highly skilled procedure such as

phacoemulsiﬁcation.

Fig. 15.4 Capsulorrhexis simulation in cataract surgery us-

ing virtual reality and haptics

In summary, all of the current paradigms available are

suboptimal from the perspective of surgical education

as well as patient care, and developments in virtual-

reality simulationwill allow the surgeon to gain realistic

operative experience before their ﬁrst cataract surgery.

The ophthalmic development of virtual-reality pro-

cedures has been minimal. The EYESI ophthalmic

surgical simulator (VRmagic GmbH, Mannheim, Ger-

many) simulates intraocular surgery but they have

a heavy reliance on physical prototyping that is hard to

reconﬁgure.

The various steps in cataract surgery can be broken

down into many steps, including incision, performing

capsulorrhexis, phacoemulsiﬁcation of the cataract and

removal, removal of cortex, and placement of the in-

traocular lens. Many of these steps are being simulated

by us using virtual reality and haptics (Fig.15.4).

15.3.3 Dental Virtual Automation

Dentistry provides a rich collection of clinical proce-

dures which require dexterity in tactile feedback and

thus provide a number of very interesting simulation

problems. The rationale for using dental simulators

is that, by using them, preclinical dental students

learn procedure skills faster and more efﬁciently. The

majority of current dental simulators use realistic

manikins along with dentiform models (Kavo, Adec

or Nevin) incorporated into a simulated dental oper-

atory [15.19]. A number of dental schools also use

the DentSim simulator (DenX Ltd., Jerusalem, Israel).

Part B 15.3

276 Part B Automation Theory and Scientiﬁc Foundations

Periodontal

probe

Periodontal probe template

Fig. 15.5 Periodontal simulations

This is a more sophisticated manikin simulator incorpo-

rating computer-aided audiovisual simulations having

a VR component. It uses a tracking system to trace the

movements of a handpiece and scores the accuracy of

a student’s cavity preparation in a manikin’s synthetic

tooth.

The addition of haptics to dental simulators is ex-

tremely important because sensory motor skills must

be developed by dental and hygiene trainees in order

to be successful in their profession. This is especially

true in the ability to perform a variety of periodontal

procedures (i.e., scaling and root planning, periodon-

tal probing, and the use of a periodontal explorer).

These tactile skills are currently acquired by having

trainees observe instructor demonstrations of a speciﬁc

procedure and then having them practise on manikins

or animal heads. After sufﬁcient training and prac-

tice on models, the students proceed to patients. This

time-consuming teaching process requires excessive

one-on-one instructor–student interaction without stu-

dents actually feeling what the instructor feels or being

physically guided by the instructor while performing

a procedure on a manikin or a patient. With the current

critical shortage of dental faculty, the training problem

has been compounded.

Dental simulators that have been developed include

those from Novint Technologies (Novint, Albuquerque,

NM) and Simulife Systems (Simulife Systems, Paris,

France). Our work has led to PerioSim at the University

of Illinois at Chicago, College of Dentistry [15.19,20].

The PerioSim prototype system for periodontal prob-

ing is designed to provide a haptic force feedback and

guidance system for interaction with an on-screen dis-

play. The display is a 3-D VR model of a periodontal

probe and a human upper and lower dental arch, which

includes teeth,their supportive structures, and other oral

tissues. The ﬁdelity of the haptic-based system is cur-

rently sufﬁciently sophisticated to differentiate between

hard and soft, and normal and pathological tissues.

While the development is taking place the other goals

are to determine how realistic the PerioSim simulator is

and the time required for learning how to use it.

The steps followed by a trainee are as follows: Us-

ing a control panel, one of threeperiodontal instruments

can be selected for on-screen use: a periodontal probe,

a periodontal explorer or a Gracy scaler. The monitor

graphic display is used in conjunction with a haptic de-

vice (PHANToM from SensAble Corp., Woburn, MA)

for force feedback and perceiving the textural feel of

the gingival crevice/pocket area. The 3-D VR periodon-

tal probe can be used to locate and measure crevice or

pocket depths around the gingival margins of the teeth.

These sites can be identiﬁed via the haptic feedback ob-

tained from the PHANToM stylus with or without the

actual instrument attached to the stylus. A trainee will

be able to differentiate the textural feel of pocket ar-

eas and locate regions of subgingival calculus. Since the

root surface iscoveredby gingiva, thetrainee cannotsee

the area being probed or the underlying calculus and

must depend totally on haptic feedback to identify these

areas. This situation corresponds to conditions encoun-

tered clinically. The control panel, which can be made

to appear or disappear as needed, has a variety of con-

trols, including adjusting the haptic feel and the degree

of transparency of the gingiva, roots, crowns, bone or

calculus. The control panel also permitsthe instructor to

insert a variety of templates to guide student instrument

positioning in a 3-D display environment (Fig.15.5).

15.4 Conclusions and Emerging Trends

Virtual-reality and automation technologies are aimed

at reducing the time spent in providing service and de-

sired training. At present almost all major companies

operate globally. Management, control, and service of

overseas facilities and products deployed overseas is

a major challenge. Both processes and factories need

to be simulated, although the current thrust seems to

be on individual process simulation. One of the impor-

tant emerging themes is constant evolution of enabling

technologies.

Part B 15.4

Virtual Reality and Automation References 277

A review ofongoing evolution of enabling technolo-

gies is available in Burdea and Coiffet’s book on VR

technologies [15.21]. The two main areas in this regard

are hardware and software technologies (e.g., [15.22]).

The hardware can be broken down into input and

output devices and computing hardware architectures.

Input devices consist of various trackers; mechanical,

electromagnetic, ultrasonic, optical, and hybrid inertial

trackers are covered. Tracking is used to navigate and

manipulate in a VR environment. There are many more

three-dimensional navigation devices that are available,

including hand and ﬁnger movement tracking. Output

devices address graphics displays, sound, and haptic

feedback. The graphics displays include head-mounted

displays, binocular-type hand-supported displays, ﬂoor-

supported displays, desktop displays, and large displays

based on large monitors and projectors supporting

multiple participants simultaneously. The haptics tac-

tile feedback covers tactile mouse, touch-based glove,

temperature-feedback glove, force-feedback joysticks,

and haptic robotic arms. Hardware architectures include

the two major rendering pipelines: graphics and hap-

tics. Personal computer (PC) graphics accelerator cards

such as those from nVidia are currently in vogue. Vari-

ous distributed VR architectures addressing issues such

as graphics and haptics pipelines synchronization, PC

clusters for tiled visual displays, and multiuser shared

virtual environments are equally important. Software

challenges include those in modeling and VR pro-

gramming. Modeling addresses geometric modeling,

kinematics modeling, physical modeling, and behav-

ior modeling. VR programming will continue to be

aided by evolution of more powerful toolkits. The en-

abling technologies outlined above can provide answers

to many of these challenges in virtual reality and au-

tomation.

References

15.1 B. Delaney: The Market for Visual Simula-

tion/Virtual Reality Systems, 6th edn. (Cyberedge

Information Services, Mountain View 2004),

www.cyber-edge.com

15.2 H.Y. Kan, V.G. Duffy, C.-J. Su: An Internet vir-

tual reality collaborative environment for effective

product design, Comput. Ind. 45(2), 197–213 (2001)

15.3 M. Pouliquen, A. Bernard, J. Marsot, L. Chodorge:

Virtual hands and virtual reality multimodal plat-

form to design safer industrial systems, Comput.

Ind. 58(1), 46–56 (2007)

15.4 B. Stone, G. Pegman: Robots and virtual reality in

the nuclear industry, Serv. Robot 1(2), 24–27 (1995)

15.5 P.R. Chakraborty, C.J. Bise: Virtual-reality-based

model for task-training of equipment operators in

the mining industry, Miner. Res. Eng. 9(4), 437–449

(2000)

15.6 S. Ottosson: Virtual reality in the product de-

velopment process, J. Eng. Des. 13(2), 159–172

(2002)

15.7 Y. Jun, J. Liu, R. Ning, Y. Zhang: Assembly process

modeling for virtual assembly process planning,

Int. J. Comput. Integr. Manuf. 18(6), 442–451

(2005)

15.8 D. Lee, M. Woo, D. Vredevoe, J. Kimmick,

W.J. Karplus, D.J. Valentino: Ophthalmoscopic ex-

amination training using virtual reality, Virtual

Real. 4(3), 184–191 (1999)

15.9 C.H.Park,G.Jang,Y.H.Chai:Developmentofavir-

tual reality training system for live-line workers,

Int. J. Human-Comput. Interact. 20(3), 285–303

(2006)

15.10 V.S. Pantelidis: Virtual reality and engineering ed-

ucation, Comput. Appl. Eng. Educ. 5(1), 3–12 (1997)

15.11 Y.S. Shin: Virtual reality simulations in Web-based

science education, Comput. Appl. Eng. Educ. 10(1),

18–25 (2002)

15.12 T.M. Rhyne: Going virtual with geographic in-

formation and scientiﬁc visualization, Comput.

Geosci. 23(4), 489–491 (1997)

15.13 O.N. Kwon, S.H. Kim, Y. Kim: Enhancing spatial vi-

sualization through virtual reality (VR) on the Web:

software design and impact analysis, J. Comput.

Math. Sci. Teach. 21(1), 17–31 (2002)

15.14 P. Queau: Televirtuality: the merging of telecom-

munications and virtual reality, Comput. Graph.

17(6), 691–693 (1993)

15.15 M. Torabi: Mobile virtual reality services, Bell Labs

Tech. J. 7(2), 185–194 (2002)

15.16 P. Banerjee, D. Zetu: Virtual Manufacturing (Wiley,

New York 2001)

15.17 P. Banerjee, C. Luciano, L. Florea, G. Dawe: Com-

pact haptic and augmented virtual reality system,

US Patent Appl. No. 11/338434 (2006), (previous

version: C. Luciano, P. Banerjee, L. Florea, G.

Dawe: Design of the ImmersiveTouch: A High-

Performance Haptic Augmented Virtual Reality

System, CD ROM Proc. Human-Comput. Interact.

(HCI) Int. Conf. (Las Vegas 2005))

15.18 P.P. Banerjee, C. Luciano, G.M. Lemole Jr, F.T. Char-

bel, M.Y. Oh: Accuracy of ventriculostomy catheter

placement on computer tomography data using

head and hand tracked high resolution virtual

reality, J. Neurosurg. 107(3), 515–521 (2007)

Part B 15

278 Part B Automation Theory and Scientiﬁc Foundations

15.19 C. Luciano: Haptics-based virtual reality Periodon-

tal Training Simulator. Ph.D. Thesis (University of

Illinois, Chicago 2006)

15.20 A.D. Steinberg, P. Banerjee, J. Drummond, M. Ze-

fran: Progress in the development of a hap-

tic/virtual reality simulation program for scaling

and root planing, J. Dent. Educ. 67(2), 161 (2003)

15.21 G. Burdea, P. Coiffet: Virtual Reality Technology,

2nd edn. (Wiley Interscience, New York 2003)

15.22 C. Luciano, P. Banerjee, G.M. Lemole, J. Char-

bel, F. Charbel: Second generation haptic ven-

triculostomy simulator using the immersivetouch

system, Proc. 14th Med. Meets Virtual Real. (2006)

pp. 343–348

Part B 15

279

Automation o

16. Automation of Mobility and Navigation

Anibal Ollero, Ángel R. Castaño

This chapter deals with general concepts on the au-

tomation of mobility and autonomous navigation.

The emphasis is on the control and navigation of

autonomous vehicles. Thus, after an introduction

with historical background and basic concepts, the

chapter brieﬂy reviews general concepts on vehi-

cle motion control by using models of the vehicle,

as well as other approaches based on the informa-

tion provided by humans. Autonomous navigation

is also studied, involving not only motion plan-

ning and trajectory generation but also interaction

with the environment to provide reactivity and

adaptation in the autonomous navigation. These

interactions are represented by means of nested

loops closed at different frequencies with different

bandwidth requirements. The human interactions

at different levels are also analyzed, taking into

account transmission of control commands and

feedback of sensory information. Finally, the

16.1 Historical Background ........................... 279

16.2 Basic Concepts...................................... 280

16.3 Vehicle Motion Control .......................... 283

16.4 Navigation Control

and Interaction

with the Environment........................... 285

16.5 Human Interaction ............................... 288

16.6 Multiple Mobile Systems........................ 290

16.7 Conclusions.......................................... 292

References .................................................. 292

chapter studies multiple mobile systems by

analyzing coordinated navigation of multiple au-

tonomous vehicles and cooperation paradigms for

autonomous mission execution.

16.1 Historical Background

The ambition to emulate the motion of living things has

been sustained throughout history, from old automata

to humanoid robots today. The Ancient Greeks and

also other cultures from around the world created many

automata and managed to make animated statues of me-

chanical people, animals, and objects (e.g., the moving

stone statues made by Daedalus in 520 BC, the ﬂying

magpie created by King-shu in 500BC, the mechani-

cal pigeon made by Archytas of Tarentum in 400BC,

the singing blackbird and other ﬁgures that drank and

moved created by Ctesibius in 280BC, etc.). In the 14th

century, remarkable automata were produced in Europe

(by Johannes Müller and Leonardo da Vinci). The de-

scriptions of mechanical automata in the 16th, 17th,

and 18th centuries are also well known (Gianello Della

Tour, Salomon de Caus, Christiaan Huygens, Jacques

de Vauc¸anson, etc.). They had complex mechanisms

and produced sounds and music synchronized with the

motions. Many interesting automata were made in the

19th century. This was the time when production tech-

niques decreased in cost. The robots (R.U.R. (Rossum’s

Universal Robots), 1921) gave their name to many me-

chanical men built in the 1920 and 1930s to be shown

in ﬁlms or fairs or to demonstrate remote control tech-

niques, such as the Westinghouse robots in the 1930s.

By the end of the 19th and beginning of the 20th

century, mass industrial production required the de-

velopment of automation. Mobility automation also

played an important role. Then, for example, primi-

tive conveyor belts were used in the 19th century and

the technology was introduced into assembly lines by

1913 in Ford Motor Company’s factory. Automation

Part B 16

280 Part B Automation Theory and Scientiﬁc Foundations

of mobility in general became a key issue in fac-

tory automation. The ﬁrst programmable paint-spraying

mechanism was designed in 1938. Industrial automa-

tion met robotics with the design in 1954 of the ﬁrst

programmable robot by Devol, who coined the term

universal automation. General Motors applied the ﬁrst

industrial robot on aproduction linein 1962.From 1966

through 1972, the Artiﬁcial Intelligence Center at SRI

International (then the Stanford Research Institute) con-

ducted research on the mobile robot system Shakey.

Control and automation in the transportation of

goods and people has also been the target for many cen-

turies. Rudimentary elevators, operated by animal and

human power or by water-driven mechanisms, were in

use during the Middle Ages and can be traced back to

the third century BC. The elevator as we know it to-

day was ﬁrst developed during the 19th century and

relied on steam or hydraulic plungers for lifting capa-

bility under manual control. Thus, the valves governing

the water ﬂow were manipulated by passengers using

ropes running through the cab, a system later enhanced

with the incorporation of lever controls and pilot valves

to regulate cab speed.

Control has also been a key concept in the devel-

opment of vehicles. Watt’s steam engines controlled by

ball governor in 1787played acentral role in the railway

developed by 1804 in Great Britain. Control also played

a decisive role in the manned airplane ﬂights of the

Wright brothers in 1903. Automatic feedback control

for ﬂight guidance was possible by 1903 thanks to the

gyroscope. Spinning gyroscopes were ﬁrst mounted in

torpedoes. The gyroscope resisted any change in direc-

tion by controlling the rudder, automatically correcting

any deviation from a straight course. By 1910 gyro-

scopic stabilizing devices had been mounted in ships,

and even in an airplane.

16.2 Basic Concepts

Automation of mobility can be examined by consider-

ing the degree of ﬂexibility. Railways are mechanically

constrained by the rails. Thus, railway automation has

been accomplished since the 1980s in France, where

automated subway lines have been in operation since

the 1990s for mass transit. The same happens with

elevators and industrial automated warehouse systems

based on overhead trolleys, as shown inFig.16.1 [16.1].

This is an effective solution in many warehouse sys-

tems and industry transportation in general. In other

cases the path of motion is also physically predeter-

Fig. 16.1 Car seat covers industrial automated warehouse

system based on overhead trolleys [16.1]

mined by different type of guides, such as inductive

guide wires embedded in the ﬂoor, which are used

in industrial automation to guide so-called automated

guided vehicles (AGVs). Other AGVs use guideways

painted in the ﬂoor that require maintenance to be de-

tectable by AGV optical sensors. Installation of new

AGVs or changing the pathways is time consuming

and expensive in these systems. A more ﬂexible solu-

tion consists of using beacons or marker arrangements

installed in the factory. These systems are more ﬂex-

ible but require line-of-sight communication between

devices to be mounted on the vehicles and a certain

number of devices in the industrial environment. In-

creasing ﬂexibility in industrial transportation has been

an objective for many years. The application of indus-

trial autonomous vehicles (Fig.16.2) and mobile robots

with onboard environment sensing for autonomous nav-

igation provides higher ﬂexibility but poses reliability

problems. In general the trade-off between ﬂexibility

and reliability continues to be a critical aspect in many

factory implementations.

Flexible automation of transportation in outdoor

environments poses signiﬁcant challenges because of

the difﬁculties in their conditioning and dynamic char-

acteristics. Automation of cars for transportation has

also been a research and development subject since

the 1990s. The autonomous navigation of cars and

vans has been a testbed for perception, planning, and

Part B 16.2

Automation of Mobility and Navigation 16.2 Basic Concepts 281

control techniques for almost 20 years. Thus, for ex-

ample, autonomous visual-based car navigation was

demonstrated in Germany by the mid 1980s, and the

NavLab project was running in Carnegie Mellon Uni-

versity by the end of the 1980s [16.2](Fig.16.3).

Many demonstrations in different environments have

also been presented. However, the daily operation of

these systems is still constrained and only in opera-

tion at low speed and in short trips in restricted areas.

In [16.3] the state of the technology and future direc-

tions of these so-called cybercars are analyzed. The use

of dedicated infrastructures and the gradual transition

from driverassistance to full automation are highlighted

as realistic paths toward fully autonomous cars. Some

applications require equipped infrastructures, such as

guideways for automatic vans with magnetic tracks

integrated in the road pavement. The majority of auto-

mated highways systems need an equipped road with

an adapted architecture, i.e., the PATH (partners for

advanced transit and highways) project [16.4].

So-called unmanned vehicles serve as means of car-

rying or transporting something, but explicitly do not

carry a human being. Thus, unmanned ground vehi-

cles in the broader sense includes any machine that

moves across the surface of the ground, such as legged

machines and machines with onboard tools and robot

manipulators (Fig.16.4). In this chapter, we will con-

centrate on the control and navigation aspects of these

vehicles.

Unmanned aerial vehicles(UAVs) are self-propelled

air vehicles that are either remotely controlled by a hu-

man operator (remotely piloted vehicles (RPV)) or are

capable of conducting autonomous operations. During

recent decades signiﬁcant efforts have been devoted to

Fig. 16.3 Navlab I at Carnegie Mellon University

Fig. 16.2 Packmobile AGV, Egemin Automation

increase the ﬂight endurance, ﬂight range, and pay-

load of UAVs. Today, UAVs with several thousands of

kilometers of ﬂight range, more than 24h of ﬂight en-

durance, and more than 1000kg of payload (Fig. 16.5)

are in operation. Furthermore, autonomous airships, he-

licopters of different size (Fig. 16.6), and other vertical

take-off and landing UAVs have also been developed.

UAV technology has also evolved to increase the on-

board computation and communication capabilities.

Fig. 16.4 RAM 2 Mobile robot with manipulator at the

University of Málaga

Part B 16.2

282 Part B Automation Theory and Scientiﬁc Foundations

Fig. 16.5 Global Hawk, Northrop Grumman Corporation

The development of new navigation sensors, actua-

tors, embedded control, and communication systems

and the trend towards miniaturization point to mini-

and micro-UAVs with increasing capabilities. In [16.5]

many UAVs are presented and compared.

Autonomous underwater vehicles (AUV) and un-

derwater robotics is also a very active ﬁeld of research

and development in which many new developments

have been presented in recent years. These vehicles

are natural extensions of the well-known remotely op-

erated underwater vehicles (ROV) that are controlled

and powered from the surface by an operator/pilot

Ancient

automata

Industrial

robots

Service

and field

robots

Distributed

automation

Flexible

manufacturing

Controlled

machines

Primitive

mechanical

machines

Old vehicles

Railways

automobiles

airplanes

Unmanned

vehicles

Intelligent

autonomous

vehicles

Teams

swarming

Motion

control

Planning

Motion

control

Planning

Environment

perception

Environment

perception

Communication Communication

Fig. 16.7 Evolution of automation in mobility and navigation

Fig. 16.6 Helicopter UAV at the University of Seville

via an umbilical cable and have been used in many

applications.

Finally, it should be noted that in many cases

mobility is strongly constrained by the particular char-

acteristics of the environment. Thus, for example, the

internal inspection [16.6] and eventual repairing of

pipes impose constraints on the design of the robots that

should navigate inside the pipe, carrying cameras and

other sensors and devices. However, these applications

will be not considered in this chapter.

The autonomy of all of the above-mentioned ve-

hicles is based on automatic motion control. The next

section will summarize general concepts of vehicle

Part B 16.2

Automation of Mobility and Navigation 16.3 Vehicle Motion Control 283

motion control. Autonomous operation also requires

environment perception. General concepts on environ-

ment perception and reactivity will be examined in the

fourth section of this chapter. Another approach re-

lated to mobility enhancement is human augmentation,

the objective of which is to augment human capabili-

ties by means of motion-controlled devices interfacing

with the human. Two different approaches can be fol-

lowed. The ﬁrst is to provide teleoperation capabilities

to control the motion of remote vehicles or devices for

transportation. The second approach is to augment the

person’s physical abilities by means of wearable de-

vices or exoskeletons. The idea is not new, but recent

progress in sensing human body signals and embedded

control systems makes it possible to build exoskele-

tons to carry heavy loads and march faster and longer.

The applications are numerous andinclude handicapped

people, military, and others. Section 16.5 will be de-

voted to the human–machine interaction for mobility

and particularly for vehicle guidance. Finally, in recent

years, a general emerging trend is the development of

ﬂeets or teams of autonomous vehicles. This trend in-

volves ground, aerial, and aquatic ﬂeets of vehicles.

Section 16.6 is devoted to the coordination of ﬂeets of

vehicles. In all the following sections the background

and signiﬁcance of each topic, as well as the existing

methods and trends are described. Figure 16.7 shows

the evolution of vehicles, robotics, and industrial au-

tomation related to the automation of mobility and

navigation, illustrating the impact of motion control,

motion planning, environment perception, and commu-

nication technology.

16.3 Vehicle Motion Control

The lowest-level vehicle control typically consists of

the vehicle motion axis in which conventional lin-

ear proportional–integral-derivative (PID) motion con-

trollers are usually applied. Other techniques such as

fuzzy logic [16.7] are less commonly used. The ref-

erence signals of these servo controllers are provided

by navigation control loops in which the objective is

that the vehicle follows previously deﬁned trajecto-

ries. In these control loops navigation sensors such as

gyroscopes, accelerometers, compass, and global po-

sitioning systems (GPS) are used. The objectives are

perturbation rejection and improving of dynamic re-

sponse. The kinematics and dynamic behavior of the

vehicle play an important role in these control loops.

The usual formulation is given by the equation

x = f (x, u) ,

(16.1)

where x ∈ R

is the vehicle state vector, usually con-

sisting of the position, the Euler angles, the linear

velocities,and theangular velocities, and u ∈

are the

control variables. In general the above equationinvolves

the rigid-body kinematics and dynamics, the force and

moment generation, and the actuator dynamics. Kine-

matics is usually considered by means of nonlinear

transformations between the reference systems associ-

ated to thevehiclebody and theglobal reference system.

The nonholonomic constraints, φ(

x, x) =0, where φ is

a nonintegral function, restrict the vehicle’s admissi-

ble directions of motion and make the control problem

more difﬁcult. Thus, it has been shown [16.8]thatit

is not possible to stabilize a nonholonomic system to

a given set-point by a continuous and time-invariant

feedback control. Moreover, in many cases, there are

underactuated control systems in which the number of

control variables m is lower than the number of degrees

of freedom. The dynamic model involves the relation

between forces and torques, generated by the propul-

sion system and the environment, and the accelerations

of the vehicle. These relations can be described by

means of the Newton–Euler equations. Furthermore, in

aerial vehicles, aerodynamics plays a signiﬁcant role

and should be considered.

In ground vehicles the tires–terrain interactions cre-

ate signiﬁcant complexity. Several models have been

developed to consider these interactions [16.9,10]. The

complexity of the complete dynamic models means that

there are a lot of vehicle, road, and tire parameters to es-

timate and tune. Besides, these models change with tire

pressure, road surface and conditions, vehicle weight,

etc. so they aremore commonlyused forcontroller anal-

ysis and simulation, braking controller research or tire

failure detection [16.11,12].

In many formulations only simpliﬁed dynamic mod-

els are considered and then only the position and angles

are included in the state vector of the above equation.

Furthermore, if the vehicles move in a plane and the

interactions with the terrain are neglected, then only

the position in the plane and the orientation angle are

considered as state variables. Sometimes a simple actu-

ator dynamic model [16.13] is added to these simpliﬁed

models. Control theory has been extensively applied

to vehicle control. Many linear and nonlinear control

Part B 16.3

284 Part B Automation Theory and Scientiﬁc Foundations

systems have been proposed to maintain stability and

tracking paths or time trajectories [16.14]. The limita-

tion in the stabilization of nonholonomic system has

been avoided by means of time-varying state feedback

and discontinuous feedback. The trajectory tracking can

be formulated by deﬁning the reference model

= f(x

, u

) . (16.2)

Then, if the model of the vehicle is (16.1), the problem

is to ﬁnd a control law u =ϕ(x, x

, u, u

) such that

lim

t→∞

x(t)−x

(t)

=0 . (16.3)

The path tracking problem consists of the tracking of

a deﬁned path, taking into account the vehicle motion

constraints deﬁned by the equation above. Different

trajectory and path tracking methods for ground ve-

hicles have been formulated by implementing linear

and nonlinear control laws, as shown and summarized

in [16.14,15]. Many practical implementations of path

tracking are based on geometric methods [16.16–18];

for example, the simple pure-pursuit method is a lin-

ear proportional control law of the error of the vehicle

position with respect to a goal point. The gain is

given by the distance to this goal point deﬁned in the

path. The appropriate selection of this gain is a crit-

ical issue. Then, gain self-scheduling approaches can

be applied. The stability of the path tracking con-

trol loop is analyzed in [16.19]. Even though these

geometric methods are not optimal, they are usually

easy to tune and have a good trade-off between per-

formance and simplicity in real-time implementations

as shown in the Defense Advanced Research Projects

Agency (DARPA) Grand Challenge 2005 competi-

tion [16.20], where most of the teams used simple

geometric approaches. Also traditional automatic con-

trol techniques such as generalized predictive control

(GPC), robust control [16.21]orLQG/LQR (linear

quadratic Gaussian/linear quadratic regulator) [16.22]

have been successfully applied. These methods require

a linear vehicle model so a simpliﬁed linearized ver-

sion of more complex models, such as the previously

commented ones, are typically employed.

The steering of ground articulated vehicles and

tractor–trailer systems (see, for example, [16.23, 24])

also presents interesting control problems. Tractor–

trailer systems (Fig. 16.8) have as an additional state

variable the angle between the tractor and the trailer.

They are usually underactuated and the saturation in the

actuators can play a signiﬁcant role, particularly when

maneuvering backwards because the vehicle tends to

Fig. 16.8 Romeo 4R with trailer in a parking manoeuvre

reach a so-called jack-knife when the angle between

the tractor and the trailer is greater than a certain value

due to perturbations coming from the vehicle–terrain

interaction.

In aerial and underwater vehicles the navigation

problem is typically formulated in three-dimensional

(3-D) space and different control levels involving the

dynamic behavior of the vehicle can be identiﬁed. The

objective of the lower levels is to keep the vehicle in

a given attitude by maintaining the stability. The lin-

ear and angular velocities in the vehicle body axis, and

orientation angles are usually considered as state vari-

ables. The higher-level motion control loop consists of

the path or trajectory tracking (which includes preci-

sion timing). In this case, the course angle error and the

cross-track distance can be considered as error signals

in the guidance loop.

In [16.25] control techniques for autonomous aerial

vehicles are reviewed. The navigation of these UAVs

is based on GPS positioning but visual-based position

estimation have also been applied [16.26], partic-

ularly in case of GPS signal loss (see the next

section). In [16.27] model-based control techniques for

autonomous helicopters are reviewed and different ex-

periments involving dynamic nonlinear behaviors are

presented. The position and orientation of an helicopter

is usually controlled by means of ﬁve control inputs:

the main rotor collective pitch, which has a direct ef-

fect on the helicopter height; the longitudinal cyclic,

which modiﬁes the helicopter pitch angle and the lon-

gitudinal translation; the lateral cyclic, which affects

the helicopter roll angle and the lateral translation;

the tail rotor, which controls the heading of the he-

licopter and compensates the antitorque generated by

the main rotor; and the throttle control. It is a mul-

Part B 16.3