This book is really unique in the robotics literature in that it presents, in a unified framework, the problems of designing, modeling, and experimenting with legged robots, a very underdeveloped field in robotics. The presentation is clear, detailed (when necessary) and sometimes done with humor. The figures are generally good and abundant. Moreover, there is a short summary at the end of each chapter that readers can use to select material of their interest, as well as to get an overview of the content of the book.

The Preface gives a short history about how and when the author got his interest in legged locomotion. His personal participation characterizes the whole book.

The first chapter introduces legged locomotion and why it is important to develop this kind of locomotion. Useful machines can be built to walk on difficult terrain, and working models can enable a better understanding of human locomotion. Animal and human locomotion have been studied by biologists and bioengineers, but, according to Raibert, the control principles that are used to walk and balance are still largely unknown. Dynamics is of primary importance in the study of legged locomotion because the velocities and masses are significant and because active balance, which allows legged systems to depart from static equilibrium positions, depends on dynamic considerations too.

An overview of previous research on legged machines introduces the main topic of this book, which is how to model and to control a legged, running machine. The theory developed from some experimental machines built by Raibert at Carnegie-Mellon University. The first was a one-legged machine, moving in a plane. Then the control algorithm was extended for movements in the 3D space. The problem of controlling a machine with several legs was approached with the same control algorithm.

Chapter 2 explains bidimensional hopping, and a planar, one-legged, hopping machine is illustrated. The control problem for this machine is decomposed into three parts: hopping, forward speed, and body attitude. These three parts of the control system are then synchronized by a finite state machine.

The control of the hopping height is based on the observation that there is an alternation between a phase in which the leg supports the load of the body and a phase in which the leg is unloaded and free to move.

The control system manipulates the accelerations to control forward speed by choosing the position for the foot before landing. The problem of computing the foot position could be solved if the equations for motion have analytic solutions. Two other solutions are discussed in the book: one is based on tabulation and is described in Chapter 7; the other is based on approximations, and has proven to be effective. To control the body attitude, the system servos move the body to an upright posture when the foot is in place during stance. Experiments with an implementation of the control system are described.

Chapter 3 develops hopping in three dimensions. The experimentation made with a new legged system has been based on extending the planar motion algorithm, in particular preserving the decomposition of the control into three parts. The kinematic model of the machine is also completely described in the Appendix provided at the end of the chapter.

In Chapter 4, the study is extended to systems with two or four legs. The starting point is that the running of bipeds is very similar to the running of a one-legged machine when the running gait uses one leg of support at a time. Therefore, the three algorithms used for controlling the one-legged machine have been used also for a two-legged machine. A more complex sequencing mechanism has been added. Control systems for n-legged systems can be obtained with the same algorithm if every couple of legs, acting in unison, is substituted by a virtual leg. The equations for virtual legs and the kinematic model of a four-legged machine are given in the Appendices at the end of the chapter.

Symmetry in running is the topic of Chapter 5. To be able to run at a fixed speed and to maintain the body in a stable upright position, the control system should be able to choose motions of the legs that give odd or even symmetry, in the appropriate combination, to the forward position and the pitch angle of the body and to the vertical position of the body. Symmetry is described as a sufficient condition, not a necessary one, to obtain functions that integrate to zero and that can be used to describe some aspects of the animal’s running, too. Equations for motions for planar systems and proof of symmetric leg motions are given in Appendix 5.

Chapter 6 introduces a few of the many alternatives for locomotion control that are available. Raibert also describes some experimentations.

Tabulation as a method for solving the motion equations is described in Chapter 7. The solutions provided make use of multivariate polynomials to approximate the tabular data and to reduce the required storage. Performance index minimization is given in the Appendix.

Chapter 8 “ties results from the study of legged machines to the study of animal locomotion.” Experiments on animal locomotion are suggested, to see whether or not the simplifications introduced in the legged robot apply to the natural systems too. The second half of the chapter explains what features are necessary in order to have a useful legged robot.

A Bibliography of more than 300 interdisciplinary references, followed by the Index, concludes the book.

The purpose of the book is to develop a scientific field, and to report on the experimental activity that accompanied the theoretical developments in legged locomotion. The book deserves attention from a broad audience that includes researchers in various interdisciplinary fields as automation, bioengineering, and robotics, in particular. Even though it is not organized as a textbook, students could benefit from it for approaching the field of legged robots (and their control problems), for which no other books are available.