Bipedal Robots E-Book

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 371

Veröffentlichungsjahr: 2013

Beschreibung

This book presents various techniques to carry out the gait modeling, the gait patterns synthesis, and the control of biped robots. Some general information on the human walking, a presentation of the current experimental biped robots, and the application of walking bipeds are given. The modeling is based on the decomposition on a walking step into different sub-phases depending on the way each foot stands into contact on the ground. The robot design is dealt with according to the mass repartition and the choice of the actuators. Different ways to generate walking patterns are considered, such as passive walking and gait synthesis performed using optimization technique. Control based on the robot modeling, neural network methods, or intuitive approaches are presented. The unilaterality of contact is dealt with using on-line adaptation of the desired motion.

Leseprobe

Table of Contents

Chapter 1. Bipedal Robots and Walking

1.1. Introduction

1.2. Biomechanical approach

1.3. Human walking

1.4. Bipedal walking robots: state of the art

1.5. Different applications

1.6. Conclusion

1.7. Bibliography

Chapter 2. Kinematic and Dynamic Models for Walking

2.1. Introduction

2.2. The kinematics of walking

2.3. The dynamics of walking

2.4. Dynamic constraints

2.5. Complementary feasibility constraints

2.6. Conclusion

2.7. Bibliography

Chapter 3. Design Tools for Making Bipedal Robots

3.1. Introduction

3.2. Study of influence of robot body masses

3.3. Mechanical design: the architectures carried out

3.4. Actuators

3.5. Sensors

3.6. Conclusion

3.7. Appendix

3.8. Bibliography

Chapter 4. Walking Pattern Generators

4.1. Introduction

4.2. Passive and quasi-passive dynamic walking

4.3. Static balance walking

4.4. Dynamic synthesis of walking

4.5. Walking synthesis via parametric optimization

4.6. Conclusion

4.7. Bibliography

Chapter 5. Control

5.1. Introduction

5.2. Hybrid systems and stability study

5.3. Taking into account the unilateralism of the contact constraint

5.4. Online modification of references

5.5. Taking an under-actuated phase into account

5.6. Taking the double support phase into account

5.7. Intuitive and neural network methods

5.8. Passive movements

5.9. Conclusion

5.10. Bibliography

Index

First published in France in 2007 by Hermes Science/Lavoisier entitles: Les robots marcheurs bipèdes: modélisation, conception, synthèse de la marche © LAVOISIER, 2007

First published in Great Britain and the United States in 2009 by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:

ISTE Ltd

John Wiley & Sons, Inc.

27-37 St George’s Road

111 River Street

London SW19 4EU

Hoboken, NJ 07030

UK

USA

www.iste.co.uk

www.wiley.com

© ISTE Ltd 2009

The rights of Christine Chevallereau, Guy Bessonnet, Gabriel Abba and Yannick Aoustin to be identified as the authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Cataloging-in-Publication Data

[Robots marcheurs bipèdes. English]

Bipedal robots: modeling, design and walking synthesis / edited by Christine Chevallereau … [et al.]

p. cm.

Includes bibliographical references and index.

ISBN 978-1-84821-076-9

1. Robots--Motion. 2. Walking. I. Chevallereau, Christine. II. Title.

TJ211.4.R6313 2008

629.8′932--dc22

2008035231

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library

ISBN: 978-1-84821-076-9

Chapter 1

Bipedal Robots and Walking

1.1. Introduction

Man has always been interested in the relationship between himself and the living world and, more particularly, in understanding how he is different from other animals. Since Antiquity and Aristotle, then during the Renaissance when the first studies were carried out in medicine and physiology, scientists have tried to understand the influence of bipedia on human evolution. In his 1680 publication, De Motu Animalium, Borelli [BOR 80] compared different bipedal species, analyzed the importance of pendulous movement and introduced spring-mass models in order to understand walking and running in humans and other bipedal animals.

In the 18th century, Doyon [DOY 66] built a whole ensemble of automats. More recent studies have tried to make parallels between passive human walking and walking executed by prototype robots such as Collins’ Walker [COL 05] or McGeer’s Straight-legged biped [MCG 90]. These studies can help us to have a better understanding of the laws needed for commands, stability or the generations of trajectories for future humanoid robots.

Research carried out in biomechanics has enabled us to interpret in more detail the principles of kinetic and potential energy transfer which contribute to defining walking. Tendons and muscles are particularly used during walking and running, acting as actuators but also as shock absorbers. Studies have shown that they temporarily store kinetic energy which is redistributed in the propulsion phases. The simple model of an inverted pendulum mounted on a compression spring placed in the leg can be used to describe running. These studies also show that one of the most influential parameters of stability during running is the start angle between the equivalent virtual leg and the ground [SEY 02]. This angle even seems to be an auto-stabilizing running parameter. These behavioral models have been confirmed by tests carried out at MIT on a one-legged jumping robot [RAI 86].

Biomechanics has also enabled us to design kinematic and dynamic walking and running models from biological systems used during simulations with interactions with the ground, and to determine internal forces. These models have enabled us to develop equivalent methods in the context of robotics.

The number of international research teams working on the subject is steadily rising. The disciplines interested in the theme of bipedal robots are: robotics, but also automatics, mechanics, information technology, biomechanics, medicine (with work on improving orthopedic prostheses and medical rehabilitation) and cinematography (e.g. computer animation). Studies in the field of sport have been concentrated on optimizing training and detecting the limiting constraints of articulations and tendons.

Section 1.2 will present a non-exhaustive overview of the biological and biomechanical approaches to this question. The notions of similarity, the characteristics of energy consumption and mobility are particularly focused on.

Section 1.3 concerns human walking. The structure of lower limbs and their muscles are given in detail to illustrate the specificity of human bipeds. Experimental results obtained from the capture of a walking movement will conclude this section.

Section 1.4 offers a historic overview of the different bipedal robots that have been created worldwide in the past.

Finally, the present day and future applications of robot bipedia will be presented in section 1.5.

1.2. Biomechanical approach

1.2.1. Biomechanical system: a source of inspiration

Among the ensemble of living things, only humans and birds use bipedia as a form of locomotion. Certain insects (cockroaches in flight for example), reptiles (e.g. lizards, geckos and monitor lizards) and mammals (e.g. primates, bears, mongooses and rats) use bipedia in exceptional cases. In the history of species, bipedia appeared during the Mesozoic periods among a large number of dinosaur reptiles. The anatomic structure of the legs of these animals is very similar, consisting of a thigh, a leg and a foot. Each lower limb has three main articulations: the hip, the knee and the ankle. The hip is a spherical link with three degrees of mobility. The knee is a link with one or two degrees of freedom (DoF). The ankle brings together two main mobilities. An important difference is the relative length of the foot and its average position during walking which can perceptibly modify the extreme values of joint angles.

Physiologists and anatomists have developed the laws of similarity from measurements taken from animals and humans. In the 1980s, the comprehensive work of Alexander’s team [ALE 77, ALE 00, ALE 04, ALE 05] enabled them to formulate the laws of geometric and energetic similarity for a large number of living animals. Figure 1.1 depicts, for example, the results obtained from the measurements of the length and diameter of femurs and tibias. For the femurs of primates, for example, these laws of similarity can be expressed in the following formula:

[1.1]

[1.2]

where l and d represent the length and diameter of the femur (mm), respectively, and m the mass of the body (kg). It has been convened that the laws of similarity can be defined by normalizing the sizes in question. The preceding relations therefore become:

[1.3]

[1.4]

where l*, d* and m* are the normalized length, diameter and mass, respectively. For the same species, studies have shown that they are satisfactory for a large number of bones in the skeleton.

In the same way, the respective durations of the single and double support phases and the relations giving the length and frequency of steps during walking or running have been established by researchers. Figure 1.2 show the measured values for the lengths of steps for different living bipeds The dotted vertical lines on the right-hand side of Figure 1.2 indicate the transition zone between walking and running for the bipedal animals under consideration [HIL 67].

We notice that this zone is small which implies that all bipeds modify their gaits for the same relative speed (as defined by Froude’s number) even when their size and morphology are different.

Figure 1.1.The law of similarity which defines the length and diameter of the femurs and tibias of primates according to Alexander et al. [ALE 79]

Figure 1.2.Relative length of walking and running steps for humans and different bipedal animals dependent on the relative speed of progression (according to [ALE 04]). The relative speed is the ratio of the forward speed divided by (where g is acceleration due to gravity and l is distance from the hip to ground in the upright position)

Energy consumption is a very important criterion in the design and study of bipedal robots. The study of the measurements of the energetic consumption of living things is also of evident interest. It is not easy to take measurements of consumed energy per living thing unit of time during walking or running or during any other physical activity. The measurement of the consumption of oxygen per unit of time seems to be the most representative of the total amount of consumption linked to physical activity. The studies of [MAN 80, THY 01] have provided a lot of information on this subject.

Simulation models have also been established from physiological data and have enabled us to plot the energy consumed per human during walking for different speeds and per unit of distance traveled.

Figure 1.3 shows the results obtained by Sellers et al. [SEL 04] which measure the energy consumption of a walker during 1 hour and traveling over a given distance (see graph legend) depending on the speed of walking (the tested subject remains immobile when he has traveled the given distance in less than an hour).

Figure 1.3.Energy consumed per human during walking depending on the average speed of progression and the distance traveled (according to [SEL 04])

Data recently published by Marden and Bejan [BEJ 06, MAR 05] show that the actuators used in nature and electromechanical or thermal actuators follow identical laws of similarity.

Marden has identified two family groups of actuator: the first produces linear movements (myosin, DNA polymerase muscle, linear electric motor) and develops a maximum force given by relation [1.5], whereas the second family group corresponds to rotation or beat movements [SCH 04] (insect or bird flight, swimming fish, electric or thermal rotating motors) which deliver a force given by relation [1.6]:

[1.5]

[1.6]

Figure 1.4.Mass-force relation for different types of actuator (according to [MAR 05])

Figure 1.4 shows the force produced by an actuator depending on its mass for a set of very different actuators with respect to the origin of the energy that they convert. The corpus of the different types of actuator can therefore be described by the curves in Figure 1.4. Relation [1.5] is represented by the continuous lined curve in the top graph, and the dotted lined curve in the bottom graph. Relation [1.6] is represented by the continuous lined curve in the bottom graph, and the dotted line in the top graph.

In short, biologists have put to the forefront the similarities between the locomotive behavior of animals regardless of their sizes and weights. Following measurements taken from animals of similar anatomic types but of differing sizes, Hill [HIL 50] arrived at the conclusion that for animals of similar morphology, speed is independent of sizes to within a scale factor. In this way, an animal makes movements of amplitude which are proportional to its size and at a frequency which is inversely proportional to this same size. During the study of the effects of scale on the structural adaptation of animals, three main types of similarity were brought to light.

The first is of geometric similarity, where two structures are geometrically similar if one can be obtained from the other by a uniform change in scale factor.

The second is the elastic similarity which allows deformations of the spinal cord without risking lesions [MCM 75].

The third is the dynamic similarity which includes morphological variations of animals of different sizes, the evolution of the movements and their delivered efforts. According to Alexander [ALE 83, ALE 84], two movements are dynamically similar if one can become the other by the uniform change of one or more of the three scale factors: length, time and force.

It has been demonstrated that when the forces of gravity and inertia are preponderate, two movements are dynamically similar only if they have the same Froude number:

[1.7]

where g is the acceleration of gravity, v is the average horizontal speed and l is the height of the hip from the ground in the upright position. Man moves from walking to running at a Froude number of 0.6 [BRU 98].

1.2.2. Skeletal structure and musculature

Bipedia has two main gaits: walking and running. When walking, there is always one of the two locomotive limbs in contact with the ground.

When running, a grounded monopodal impulse propels the body up and forwards in the form of a leap, followed by a new grounding with a shock to the other locomotive limb.

The kinetic energy of this ballistic phase is partially recovered at renewal support, thanks to the elasticity of the articulation, muscles and tendons.

The energy necessary for the leap is partially recovered when new contact is made with the ground. There is a left to right alternate weight transfer between the lower limbs.

Monkeys only use bipedia occasionally. They are not actual quadrupeds either because their upper and lower limbs end in hands with opposable thumbs, which are adapted for brachiation. They walk in this way on four hands which, in addition, have no pedal arches.

Bears also use bipedia very occasionally. Nevertheless, they are real plantigrade quadrupeds; they do not have opposable thumbs on any of their legs and when walking, their pedal arches are fully in contact with the ground.

The locomotive system of kangaroos is made up of lower pelvic members and a tail.

Birds, whose origins can be traced back to the theropod dinosaurs of the Jurassic period [DER 70], use a terrestrial system of locomotion where only the pelvic limbs remain, dissociated from the caudal region from their distant Jurassic ancestors.

The feet of birds are characterized by a tri-segmented Z structure: the femur, the tibiotarsus and the tarsometatarsus [MED 06]. The spinal cord of birds is fixed from the pelvis to the nape. Its central skeleton can therefore be considered as an undeformable solid.

The human spinal cord has a great number of individual mobilities. The trunk is also articulated in this way. Its movements have an influence on its locomotion. The spinal cord has a system of muscular tensors which take into account the constraints due to the alternating weight transfers from the lower right to the left limb during walking or running.

The human locomotive system stores and gives back energy (see section 1.2.2) due to the elasticity of the foot’s arch, its muscles, its ankles and its spinal discs.

Nordez’s [NOR 06] work on this matter gives a very detailed characterization of the passive stretch of the musco-articular complex. It was noticed that a static stretch brings about a mainly transitory increase in the length of muscles, whereas the dissipative properties and stiffness are only modified after cyclic stretches.

Nevertheless, to maintain an adequate level of mechanical energy for locomotion, the muscles transform chemical energy into mechanical energy, resulting in a global energetic demand.

To carry out the mechanical modeling of anthropomorphic structures, some authors have suggested models for the distribution of mass, the choice of segmented joint-links (and the number of segments) and their geometry [HAN 64].

To illustrate this, Figure 1.5 shows the skeletons of two bipedal animals (a bird and iguanodon) and a human in the upright position.

Figure 1.5.Skeletons of three bipeds: (a) a bird; (b) an iguanodon; (c) a human

1.3. Human walking

1.3.1. Architecture

The architecture of a human’s lower limbs is very complex (see Figure 1.6). Its bone structure regroups 44 bones, of which the femur, tibia and fibula are the main bones. Apart from the knee-cap, the remaining bones make up the foot, which can be considered as a deformable composite corpus.

Each member therefore has an ensemble of three main corpuses (thigh, leg and foot) linked together by articulations which have many degrees of mobility.

In this way, the hip articulation has 3 revolutionary degrees of mobility. The articulations of the knees and the ankles each have two DoF.

The musculature of a human’s lower limbs is made up of 46 skeletal muscles. The muscles which intervene in the propulsion movement are longer muscles. In this way, we notice that many muscles intervene simultaneously to assist in the motion of an articulation. The muscles make an effort of traction and they always work in conjunctive pairs with another opposite muscle.

We can also separate muscles into those which only act on a single articulation (e.g. the iliacus muscle on the hip) and others, which act on separate body-links, which are separated by two articulations (e.g. the rectus femoris muscle).

Figure 1.6a.Structure of human lower limbs: skeleton of the right leg

Figure 1.6b.Structure of human lower limbs: ventral frontal view of the muscular groups

1.3.2. Walking and running trajectory data

In a healthy subject, locomotion can be reduced to a cycle [GAG 90, LAA 92, NIL 89], where the articulate sequence is repeated as long as the speed is constant. Walking, for example, is a succession of grounding phases and balance. It has been convened that the walking cycle starts when contact is made with the right heel, followed by two steps, left then right, where the walker maintains at least one grounding. By dividing up this cycle into percentages of its duration [BEA 03, VAU 84], the contact of the right heel with the ground is considered as instantaneous (0% of the duration of the cycle) (Figure 1.7).

Both feet are grounded for 0–15% of the cycle. This is a phase of double support which corresponds to the grounding reception of the right leg, and the propulsion of the left leg. In a progressive way, the left foot leaves the ground, until the big toe has left the ground. For 15–50% of the cycle, only the right foot is grounded in a monopodal phase. The left leg is in a balancing phase. When the left heel impacts with the ground and the left foot grounds, there is a second phase of double support for 50–65% of the cycle. This phase is completed by the big toe of the right foot leaving the ground. For 65–100% of the cycle, we are in the monopodal phase on the left foot. The right leg is in the balancing phase. The cycle terminates by a new grounding of the right heel.

Human walking is characterized by a phase of double support which disappears during speeds faster than 2.1 m s−1. This disappearance corresponds to the theoretical transition of walking with an instantaneous double support to running which represents 50% of the cycle (see Figure 1.7).

Figure 1.7.Grounding phases (gray areas) expressed in percentages of the walking cycle duration

According to various studies [AND 77, LAR 80], distance D which is covered during a cycle increases linearly with speed, until it reaches a maximum. Van Emmerik and Wagenaar [VAN 96] have shown a variation of 0.6–1.4 m for speeds of 0.3–1.3 m s−1. A hysteresis phenomenon has also been shown [VAN 96]. For a given velocity, D is noticeably longer when the speed is increasing than when it is decreasing.

During walking, the frequency F of a step significantly increases with the speed [CAV 86]. Thus, F varies by around 0.5–1.0 Hz for speeds from 0.3 m s−1 to 1.3 m s−1. A phenomenon of hysteresis has also been observed: for a given speed, F reaches a higher value at increasing speed than at decreasing speed.

The main articulations associated with human locomotion are those of the hips, the knees, the ankles and the metatarsal articulations. The movement of the hip combined with the rotation of the pelvis enables humans to lengthen their step [NOV 98]. During a walking cycle, the movement for the hip in the sagittal plane is essentially sinusoidal. In this way, the thigh moves from back to front and vice versa. The articulation of the knee allows for movements of flexion and extension in the leg during locomotion. As for the ankle, the movements of flexion occur when the heel is re-grounded. There is a second flexion during the balancing phase.

The results shown in Figures 1.8–1.11 are relative to measurements1 of kinematic variables for the movement of a walking male of a height of 1.7 m. The optoelectronic acquisition system used to carry out these measurements gives a three-dimensional analysis of the movements of a polyarticulated system such as the human body, equipped with reflective anatomic frames. Seven infrared cameras (which are synchronized and set to 60 Hz) give to the acquisition software [BEA 03] an image and the coordinates of each frame. Figure 1.8 shows the movements of flexion and extension according to the time measurement of the posterior surface of the left iliac spine in the sagittal plane during one of the cycles of an established walk. The sinusoidal feature of the movement of the hip articulation is easy to identify, even if there isn’t perfect symmetry between the two extensions. Figure 1.9 shows the flexion and extension movements according to the time measurement of the left knee in the sagittal plane during the same cycle. For the left leg, the indications given on the duration of the grounding and balancing phases are given in Figures 1.10 and 1.11, which show the vertical sides of the heel and big toe of the left foot.

We notice that the heel is in contact with the ground from 0 to 0.4 s and after 0.9 s. The big toe is grounded from approximately 0.35–0.6 s. There is therefore a balancing phase for the left leg for a time span of 0.6–0.9 s.

Figure 1.8.Flexion-extension of the posterior surface of the left iliac spine during a walking cycle

Figure 1.9.Flexion-extension of the left knee in the sagittal plane during a walking cycle

Figure 1.10.Evolution of the altitude of the left heel during a walking cycle

Figure 1.11.Evolution of the altitude of the left big toe during a walking cycle

A human’s center of gravity is situated at about 55% of its height from the ground. When walking, the human center of gravity oscillates both vertically, by following a path of approximate cycloidal amplitude equal to about 75 mm and horizontally, by adopting a sinusoidal amplitude equal to about 30 mm. Researchers [CAI 94] have shown that the actions of rotating the pelvis around a vertical axis, tilting the pelvis to the side of the non-carrying leg, flexing the knee when the heel grounds, moving the foot and the heel, coordinating movements of the knee and ankle and the lateral shift of the pelvis all help to stabilize walking by playing on the position of the center of gravity.

When walking, the potential and kinetic energies of the human body are in opposition phases. The level of energy is maintained due to the transfer between the potential and kinetic energies. The potential and kinetic energies reach both a maximum and minimum value during the phases of double support. The potential energy reaches its maximum value when the mass center point goes over the grounded foot [BIA 98, NOV 98]. This corresponds to the position of the highest mass center point. When in a running gait, the potential and kinetic energies are synchronized and they reach their minimum and maximum values in the middle of the grounded phase and in the middle of the airborne phase. The transfers of kinetic/potential energy are no longer possible. The level of energy is maintained due to the storage of potential energy as a result of the elastic deformation of the body parts (e.g. the pedal arch) and due to the transfer of energy of a body segment to another through muscular articulations [NOV 98]. The muscles must therefore accomplish more mechanical work during running than during walking.

1.3.3. Study cases

The design of bipedal robots and especially humanoid robots is naturally inspired from the functional mobilities of the human body. Nevertheless, the complex nature of the skeletal structure as well as the human muscular system cannot be reproduced in robotics. The number of internal mobilities is therefore limited to the essential and the actuating system must be simplified. A bipedal robot or a humanoid robot therefore has fewer DoF than a human body. The choice of the number of DoF for each articulation is very important. The approach consists of analyzing the structure of the robot from three main planes: the sagittal, frontal and transversal planes. The movement of walking mainly takes place in the sagittal plane; all bipeds have the largest number of important articulations in this plane.

Figure 1.12 shows the typical configurations of bipedal robots in the sagittal plane. The segmented bodies are usually modeled by a punctual mass placed at their center of gravity. In Figure 1.12, the articulations are located by the black circles.

The structure in Figure 1.12a corresponds to the simplest robot, which only has two articulations at the hip. Structure 1.12b has two additional articulations at the knees.

The last figure, Figure 1.12c, is more complete with two additional articulations at the ankles. This is the configuration that is most often used for the construction of bipedal robot prototypes.

Figure 1.12.Studies of configurations for bipedal robots in the sagittal plane

The analysis of articulations in the frontal plane shows their importance for the lateral stability of walking.

Figure 1.13 shows the given solutions obtained by the designers. The stability in the frontal plane depends on the position of the center of gravity in relation to the contact point of the stance leg with the ground. In this way, there are two possible structures for moving trunk mass in the frontal plane.

The structure in Figure 1.13a allows for the displacement of the trunk’s mass above the grounded leg with a lateral flexion of the trunk at the level of the pelvis-trunk articulation.

The structure of 1.13b enables it to laterally move the trunk mass due to the articulation of the hips. An independent movement of the balancing leg is equally possible due to the independent articulation of the second hip.

The structure of Figure 1.13c is more complete; it enables it to perform a lateral transfer of the trunk by combining a double movement of adduction-abduction at the hips and ankles. In addition, keeping the feet flat is guaranteed by a better adherence to the ground.

Figure 1.13.Bipedal robot study cases in the frontal plane

The articulations in the transversal plane only exist in the most complex bipeds. These robots have one to five articulations in the transversal plane. Figure 1.14 shows the possible configurations.

The structure of Figure 1.14a, with its movement of trunk rotation, enables it to compensate for the coupled reactions in the transversal plane due to the movement of the balancing leg.

The structure of Figure 1.14b enables the robot to pivot its balancing leg by a movement of internal-external rotation, which enables it to make changes in direction when walking.

Finally, the structure of Figure 1.14c is the most complete and, in addition, enables it to orientate its foot when it is about to touch the ground or to pivot on the grounded foot.

Figure 1.14.Configuration studies of bipedal robots in the transversal plane

1.4. Bipedal walking robots: state of the art

1.4.1. A brief history

Man has always been fascinated by making systems in his own image. Leonardo de Vinci is probably the first man to have drawn (and perhaps even built) a humanoid mechanism (Figure 1.15).

The 18th century was a fertile period, with the creation of many automats able to reproduce human movements when placed in specific contexts of tasks (e.g. writing or playing music).

The 19th century was a period of construction when the Steam Man (moved by steam-engine) was built by John Brainerd and the Electric Man was built by Frank Reade Junior (Figure 1.16).

At the beginning of the 20th century, the Westinghouse society made the Elektro humanoid. It was during the 1960s and 1970s that legged robots really started to appear, especially in Japan (section 1.4.2). In Russia, the University of Lomonosov in Moscow and the University of Saint Petersburg built legged robots very early on (Mascha, Rikscha and OstRover robots) [GRI 94, GUR 81]. In the USA, the first legged robot creations controlled by computers were quadrupeds and hexapods [BRO 89, OZG 84]. The hexapod CMU, built during the period 1980–1983, reached a maximum speed of 0.11 m s−1.

Figure 1.15.Leonardo de Vinci’s Humanoid

Figure 1.16.Frank Reade Junior’s Electric Man

Figure 1.17.Spring Flamingo from Gill Pratt’s MIT laboratory

Figure 1.18.3D bipedal robot from Marc Raiberts MIT laboratory

We should also mention the work carried out at MIT (Massachusetts Institute of Technology) in the 1980s [MUR 84, RAI 83, RAI 86] on jumping robots. The bipedal robots called Biped Planar, Spring Flamingo, Spring Turkey, Uniroo and 3D Biped (Figures 1.17 and 1.18) were among the first to perform walking and running movements in a dynamic and stable gait. Their dynamic performances and the variety of their tested moving gaits were remarkable.

1.4.2. Japanese studies and creations

In the 20th century, the first studies concerning bipedal robots were carried out in Japan, where certain researchers had been interested in the subject since the 1970s. The robotic team of Waseda University must be mentioned in this context as they developed a whole family of WL (Waseda Legged) robots. Figure 1.19 shows one of the latest developments of these robots, the WL-10R [TAK 84] which has 10 articulations motorized by electrical servomotors and body parts made of plastic which are reinforced with carbon fibers. This robot was able to walk both forwards and backwards and turn around, which was a real achievement in 1983. These developments were brought about by Professor Kato’s team as part of the Wabot project, who also constructed the first anthropomorphic robot which was entirely controlled by hydraulic actuators.

The Waseda university team then developed a whole variety of bipedal and humanoid robots; the most recent have 41 motorized joints. The WABIAN-2R robot [OGU 06] is one of the most accomplished examples (see Figure 1.20). It is 1.53 m in height and weighs a total of 64.5 kg (4.5 kg for the battery). Each of its legs has 7 articulations, 4 articulations placed at the pelvis and trunk and the remaining 23 in its arms and neck. The majority of the joints are activated by the Harmonic Drive gearbox, coupled with DC motors. Each foot is equipped with a force sensor with 6 components and the control is based on the ZMP feedforward drive (see Chapter 5). Its average walking speed is 0.36 m s−1 with a period of 0.96 s per step.

Among the most interesting of Japanese research, the work of Sano and Furusho’s team is of particular interest [SAN 90, SAN 91]. From 1984 to 1988 they worked on the BLR-G2 robot which had 9 DoF and was controlled by DC motors. This robot’s maximum speed of progression was 0.35 m s−1. In the same way, Kajita and Tani [KAJ 96] built the MELTRAN II robot in the 1990s, which had passive articulations at the ankles. One of its laws of control was a function which depends on the angle of the equivalent virtual leg (see Chapter 5).

Figure 1.19.(a) WL-10R Robot, 1983; (b) Wabot-1 robot, Waseda University, Tokyo, Japan, 1973

Figure 1.20.(a) BIPER-4 robot, Tokyo University, 1984; (b) WABIAN-2R, Waseda University, Tokyo, Japan, 2004

In the 1980s, Miura and Shimoyama [MIU 84] developed the bipedal robot family called BIPER which was statically unstable but which had a dynamically stable walk. The BIPER-4 robot (Figure 1.20) for example had non-motorized articulation at the ankles, very big feet and no articulation at the knees. The analogy of an inverted pendulum’s movement was used to define its gait.

In Japan, the industrial companies developed bipedal and humanoid robots very early. Honda, in particular, built a whole range of bipedal robots from 1986 onwards. First there was E0 to E6, then humanoid robots called P1 to P3 and finally, the most complete, ASIMO (se Figure 1.21). The ASIMO robot is 1.4 m high, has 26 DoF and is moved by 26 electric motors.

Figure 1.21.ASIMO humanoid robot built by HONDA, Japan

1.4.3. The situation in France

The first studies of legged robots were carried out at Strasbourg University in the LSIT laboratory, which culminated in the creation of a bipedal robot made up of six body parts but with no trunk (just legs and feet) [CHA 93]. We can also mention the work on monopodal jumping robots [FRA 98].

Another dimension was reached with the design and construction (two models were made) of the BIP2000 biped, which was jointly made by the INRIA Rhône-Alpes and the Poitiers LMS (Figure 1.22). This 3D bipedal robot is 1.8 m high and weighs 105 kg. Its locomotive system has 12 basic mobilities which enables it to perform walking gaits similar to that of a human. It also has a pelvis-trunk articulation with three DoF. It is made up of eight main body-parts and seven articulations, which adds up to a total of 15 degrees of internal mobility. Statically stable trajectories were obtained for a walking speed equal to 0.36 km h−1.

The Rabbit project, which started in 1998 with CNRS backing, enabled the French bipedal robotic community to further their studies on stability and the control of underactuated bipedal robots. The aim of this project was to obtain walking and running gaits which were dynamically stable, based on a simple mechanical anthropomorphic structure which only had a few DoF.

The project culminated in a five-bodied biped with four motors, two on the hip and two at the top of the thighs (see Figure 1.22). Each of the four gearboxes can give a maximum nominal couple of 150 Nm. This value is necessary for running gaits. The biped is maintained in the vertical position by a pivoting horizontal beam, and moves along a circular trajectory in this way. This secure layout enabled them to perform successful walking and running trials for significant distances.

The LIRIS Laboratory at the University of Versailles made an experimental anthropomorphic biped called ROBIAN (Figure 1.23) [MOH 04] (with backing from the French Ministry of Education, Research and Technology).

This biped has a three-dimensional kinematic architecture and has 16 degrees of motorized freedom. It weighs 29 kg and is 1.30 m high. The foot is made up of an articulated forefoot along a transversal axis moved with a compliant link. The trunk is made up of a mechanism which activates three mobile masses so that it can make weight transfers in three directions. The hip module is made up of a parallel kinematic system which enables the activation of its three DoF. Two linear actuators create movements of abduction-adduction and internal-external rotation. A rotating actuator enables flexion-extension.

Figure 1.22.(a) BIP robot (INRIA Rhône-Alpes, LMS Poitiers); (b) Rabbit robot (CNRS Grenoble)

In Japan, the Humanoid Robotics Project came about under the instigation of the Ministry of Economics, Commerce and Industry (METI). The project came into being with the creation of a simulation platform (OpenHRP) and the creation of a humanoid. Its second version is called HRP-2 (Figure 1.24). This humanoid was designed by the National Institute of Science and Advanced Industrial Technologies (AIST) and built by Kawada Industries. There are 15 HRP-2 models in the world. 14 are to be found in different laboratories in Japan and one is at the Analysis and Architecture of Systems Laboratory (LAAS) at the CNRS in Toulouse. The purchase of this humanoid was made within the context of the Franco-Japanese Joint Robotics Laboratory (JRL) [KHE 07]. The HRP-2 robot is 1.5 m high, weighs 58 kg and has 30 DoF. It can move at a speed of 2.5 km h−1. It has vision cameras and force and attitude sensors so that it can control its own balance, as well as plan and control its tasks.

Figure 1.23.Robian robot made by LIRIS

Figure 1.24.HRP2 Humanoid made by Kawada

1.4.4. General evolution tendencies

In the following years, the European Network of Excellence EURON carried out a very interesting study on the development of robotics. The 2005 report [EUR 05] summed up the previsions for the different types of mobile robots and detailed the future applications of service robots in the workplace and at home. The study estimated that a total of 2 million service robots of all kinds would be used in the world by the end of 2004, either for professional or domestic use. This highly increasing number could quadruple before the end of the present decade. The estimated rise is higher still for all robots destined for domestic use (robots to help with housework or robot toys). Within the category of robots for professional use, robots which help with surgery and defense robots (surveillance, de-mining and warfare) are on the highest increase. However, this report gives us no indication of the part to be played by bipedal or humanoid robots within this market.

Important projects were launched in Japan and in the USA to develop robotics and application domains. The Japanese Ministry for Economics, Commerce and Industry (METI) invested more than 17 million dollars of its 2007 budget in order to back the development of perfectly autonomous intelligent robots which could make their own decisions in the workplace. The final aim is to introduce intelligent robots to the market by 2015. The types of intelligent robots envisaged by METI included security and cleaning robots. We could simply indicate a type of task to carry out, and it would be executed in an intelligent manner.

In the USA, among the many projects underway in the different universities, the projects of the DARPA (Department of Defense Agency) are of particular interest. Indeed, many of their projects are centered on the development of autonomous machines (unmanned autonomous vehicles) to help improve transportation and exploration in unknown territories.

Legged vehicles are therefore particularly adapted for accompanying ground troops and for traveling in difficult terrain with obstacles or escarpment. Many exoskeleton projects are also underway. We should also note the very ambitious NASA project to create an astronaut robot. Indeed, the Robonaut project consists of building and testing an autonomous humanoid robot capable of helping other astronauts, taking down information and manipulating objects in a space capsule or when exploring planets.

1.5. Different applications

Robotics is a priority in technologically advanced countries. With the ageing population, the needs in terms of service and innovating technological advance create a particularly propitious context for robotics. Indeed, there are many domains of application for robotics. The 2005 report by the EURON [EUR 05] network mentions two main robot categories for application purposes: service robots for professional use and service robots for domestic use.

For professional use robotics, the applications most often used are: agricultural and environmental robots (e.g. harvesting robots, milking robots or lumberjack robots), mining company robots, hostile territory exploration (e.g. planets or underwater), environmental surveillance (forest fires), cleaning robots (e.g. floors, walls, exterior walls or pipes), inspection robots, service robots (e.g. surveillance, security, handling or logistics), building robots for contractors and demolishers, medical robots (e.g. surgery and therapy), military robots, robot for welcome desks and as guides (e.g. hotels or museums) and graphic animation for the cinema industry in particular.

In robotics for domestic use, there are service robots (e.g. carers, housework or leisure activities), robots for company, interfaces for man-machine interactions and robots for helping the handicapped. In the following section, we will present a few examples of applications that we have already mentioned.

1.5.1. Service robotics

Service robotics for companies and individuals will be on the increase in the next few years. This domain is currently dominated by wheeled robots, but some of the bigger Japanese companies have been working on many different humanoid robot projects with an end to using them as service robots for daily tasks (e.g. maneuvering or distribution), help with housework and hospital servicing work.

The most well-known projects are those of Honda and Sony as well as the HRP-2 robot which has already been mentioned. These companies clearly advertise their objectives, which are to market robots to do the housework or to help the elderly.

Bipedal robots will be improved and developed within this context, as they will be better adapted to an environment that is usually destined for humans. Moreover, their anthropomorphic aspect will ensure that they are accepted psychologically in a more spontaneous way. A robot companion may be better accepted by a person it is caring for as it cannot make a moral judgment about this.

In service robotics, the most advanced bipedal robot project concerns the elderly. Researchers at the University of Waseda have come up with a mobile chair mounted on legs (see Figure 1.25).

The KAIST research center at the HUBO Lab has created a bipedal armchair which is in complete working order (Figure 1.26). These legged armchairs have been designed to help improve the mobility of paraplegics. In 2005, Toyota presented the i-foot prototype to help people of reduced mobility.

Figure 1.25.Robot to help the elderly, Waseda University, Japan

Figure 1.26.Hubo FX-1 robotic armchair, HUBO Lab, KAIST, South Korea

1.5.2. Robotics and dangerous terrains

Robotics for dangerous environments will also be further developed. Applications in the nuclear domain [GEL 90] or for interventions in dangerous situations (accidents) are being studied by many international teams.

To offer assistance during fires, researchers first designed wheeled and tracked-type robots, or snake-like robots with a water-hose nozzle. An important project to build a robot to help firemen is currently under way in Tokyo [MIY 02]. Buildings or sites that are mainly dedicated to human activity need mobile machines which are capable of moving around rapidly [AMA 02]. Bipedal robots have the potential mobile capacity to be able to intervene in this type of environment. In the case of catastrophes, such as an earthquake, robots can also intervene to explore destroyed buildings or locate and save victims. In the case of nuclear catastrophes, the robotic machines are also capable of exploring and transmitting collected data from contaminated zones and, if necessary, getting even closer to the source of contamination [BRI 98]. Numerous projects are underway in various countries for intervention tasks or for dismantling in the nuclear industry. However, robotic interventions of this kind are currently still carried out by wheeled or tracked-type machines as their operation is sufficiently mastered and secure.

1.5.3. Toy robots and computer animation in cinema

It is possible to buy kits of little humanoid robots such as those sold by Hitec Robotics (who commercialized the Robonova I robot, Figure 1.27). It is 0.4 m high and weighs 1.6 kg. In the same vein, Sony developed the robot dog AIBO (Artificial Intelligence roBot) and marketed it for the general public. It can interact with the environment and has sound, voice and facial recognition sensors. Mini football tournaments were organized using the AIBO robots as players. Sony has recently developed a humanoid called QRIO (Figure 1.28) which is not out on the market yet.

Other robots are dedicated to computer animation for cinema and are sometimes referred to as the growing branch of robotics called “animatronics”. With an increase in the number of scenarios including special effects with prehistoric, legendary or science fiction animals, the cinema industry is constantly looking for materials and software to help them realistically reproduce the movements of these real or long-lost animals.

Figure 1.27.Robonova-I robot made by Hitec Robotics, Japan

Figure 1.28.QRIO robot made by Sony, Japan

1.5.4. Defense robotics

In the USA, the Pentagon has predicted that robots will make up a major part of the American military force in the next ten years. A budget of 127 billion dollars has been given to the project called Future Combat Systems which represents the biggest military contract in the history of the USA. The first generation of robotic systems will take the form of a remote-controlled family of rolling machines. With developments in technology, a second generation could evolve towards more diverse conformations and morphologies, including human forms, (although certain specialists think that this will not be possible for another 20–30 years). By becoming increasingly “intelligent”, these robots will also be increasingly autonomous. To military minds, these machines will be able to distinguish friend from foe and soldier from innocent citizen.

In France, the General Direction for Armament (DGA) has put together the SYRANO prototype (robotic acquisition system for neutralizing objectives). The Robosoft company also presented an all-terrain robot called Roburoc 6.