Uncertainty is a catch-all term that refers to any and all of the different ways in which the sensors, models, and actuators of a robot do not behave ideally. Sensors rarely give exact and noise-free measurements, environmental models rarely capture all information that matters, and a wheel that does not slip has never been built. As is the case for so many other human capabilities, we rarely notice our own competence dealing with noisy, missing, and conflicting information. This chapter covers some elementary concepts of random processes and then presents classical techniques that can be used to make robots more competent in our uncertain world.

Uncertainty models can be applied to many forms of error, whether the errors are random or systematic, temporal or spatial. Every measurement has some amount of random noise impressed on it as a matter of basic thermodynamics. Modelling that noise can lead to systems of considerably higher performance. Such performance improvements may lead to improved safety, availability, and robustness.

The most basic necessity of a mobile robot is the capacity to measure its own motion. For otherwise, how could it determine which way to go or whether it ever got there? In the plane, the pose of the robot is defined as the position and orientation coordinates (x, y, ψ) of some distinguished reference frame on the body. In some cases, velocities or higher derivatives or perhaps articulations are desired and this more general problem is referred to as state estimation. Determining the exact pose of the robot is not usually a practical goal. The best that can be done is to estimate pose so it is common to refer to the problem of localization as pose estimation. Much of the theory of estimation is directly applicable.

For any element of the robot pose, an important distinction relates to the question of whether available measurements project onto the pose itself or onto any of its time derivatives. In the former case, algebraic equations relate the pose to the measurements and the technique is classified as triangulation. In the latter case, differential equations relate the pose to the measurements and the technique is classified as dead reckoning. The errors that occur in these two fundamental techniques behave in very different ways.

This section describes the methods used to implement the perceptive autonomy layer that was initially described in Chapter 1. Perception is the process of understanding the environment based on measurements. Often the process involves the construction of models that are subsequently interpreted and improved or expanded. Although localization may enable mobility, it is environmental perception that enables a system to respond intelligently to what is out there – even when it differs from any expectations. Oftentimes, sufficiently intelligent behaviors emerge naturally based only on perception. Later, in Chapter 10, we will see that an ability to predict the future is sometimes necessary too.

Perception is a rapidly expanding area because it has many applications beyond robotics. For this reason, because space is limited, and because there are already many textbooks that concentrate on perception, this section will present only the bare essentials and only those that are most applicable to mobile robots.

Image Processing Operators and Algorithms

Perception and state estimation have a lot in common. Whereas state estimation estimates the state of the robot, perception estimates the state of the environment. Although states estimation tends to deal with signal variation over time, perception tends to deal with signal variation over space.

Kinematics is the study of the geometry of motion. Kinematic modelling therefore enjoys a particularly distinguished role as one of the most important analytical tools of mobile robotics. We use kinematics to model the gross motion of the robot body as well as the internal motions of mechanisms associated with suspension, steering, propulsion, implements, and sensors. Kinematic models are employed offline for such diverse purposes as design, analysis, and visualization. Such models are also used online on computers used to interface to robots or on the computers on the robot. Many kinematic models are based on matrices and we will need a few advanced topics on matrices for a few other purposes, so matrices are reviewed first. The tools developed in this section will be used throughout the rest of the book.

Conventions and Definitions

This chapter will present some aspects of notation that will be unfamiliar to most readers who have not studied advanced dynamics. We do not need advanced dynamics in this book but we do need precise notation. We also need notations powerful enough to handle many concepts from dynamics rather than just homogeneous coordinates applied to points and vectors in space. For these reasons, we will depart from the conventions of many other robotics books and use the notations of physics rather than those of computer graphics. The reader is advised to skim this subchapter first up to Figure 2.7 and the commentary afterward. With that context, a second reading would be more illuminating.

MULTIBODY SYSTEMS

The primary purpose of this book is to develop methods for the dynamic analysis of multibody systems that consist of interconnected rigid and deformable components. In that sense, the objective may be considered as a generalization of methods of structural and rigid body analysis. Many mechanical and structural systems such as vehicles, space structures, robotics, mechanisms, and aircraft consist of interconnected components that undergo large translational and rotational displacements. Figure 1.1 shows examples of such systems that can be modeled as multibody systems. In general, a multibody system is defined to be a collection of subsystems called bodies, components, or substructures. The motion of the subsystems is kinematically constrained because of different types of joints, and each subsystem or component may undergo large translations and rotational displacements.

Basic to any presentation of multibody mechanics is the understanding of the motion of subsystems (bodies or components). The motion of material bodies formed the subject of some of the earliest researches pursued in three different fields, namely, rigid body mechanics, structural mechanics, and continuum mechanics. The term rigid body implies that the deformation of the body under consideration is assumed small such that the body deformation has no effect on the gross body motion. Hence, for a rigid body, the distance between any two of its particles remains constant at all times and all configurations. The motion of a rigid body in space can be completely described by using six generalized coordinates.

In the classical finite-element formulation for beams and plates, infinitesimal rotations are used as nodal coordinates. As a result, beams and plates are not considered as isoparametric elements. Rigid body motion of these non-isoparametric elements does not result in zero strains and exact modeling of the rigid body inertia using these elements cannot be obtained. In this chapter, a formulation for the large reference displacement and small deformation analysis of deformable bodies using nonisoparametric finite elements is presented. This formulation, in which infinitesimal rotations are used as nodal coordinates, leads to exact modeling of the rigid body dynamics and results in zero strains under an arbitrary rigid body motion. It is crucial in this formulation that the assumed displacement ield of the element can describe an arbitrary rigid body translation. Using this property and an intermediate element coordinate system, a concept similar to the parallel axis theorem used in rigid body dynamics can be applied to obtain an exact modeling of the rigid body inertia for deformable bodies that have complex geometrical shapes. More discussion on the use of the parallel axis theorem in modeling the inertia of rigid bodies with complex geometry is presented in Chapter 8 of this book.

Thus far, only the dynamics of multibody systems consisting of interconnected rigid bodies has been discussed. In Chapter 2, methods for the kinematic analysis of the rigid frames of reference were presented and many useful kinematic relationships and identities were developed. These kinematic equations were used in Chapter 3 to develop general formulations for the dynamic differential equations of motion of multi-rigid-body systems. In rigid body dynamics, it is assumed that the distance between two arbitrary points on the body remains constant. This implies that when a force is applied to any point on the rigid body, the resultant stresses set every other point in motion instantaneously, and as shown in the preceding chapter, the force can be considered as producing a linear acceleration for the whole body together with an angular acceleration about its center of mass. The dynamic motion of the body, in this case, can be described using Newton-Euler equations, developed in the preceding chapter.

In recent years, greater emphasis has been placed on the design of high-speed, lightweight, precision mechanical systems. These systems, in general, incorporate various types of driving, sensing, and controlling devices working together to achieve specified performance requirements under different loading conditions. In many of these industrial and technological applications, systems cannot be treated as collections of rigid bodies and the rigid body assumption is no longer valid. In such cases, a mechanical system can be modeled as a multibody system that consists of two collections of bodies.

While a body-fixed coordinate system is commonly employed as a reference for rigid components, a floating coordinate system is suggested for deformable bodies that undergo large rotations. When dealing with rigid body systems, the kinematics of the body is completely described by the kinematics of its coordinate system because the particles of a rigid body do not move with respect to a body-fixed coordinate system. The local position of a particle on the body can then be described in terms of fixed components along the axes of this moving coordinate system. In deformable bodies, on the other hand, particles move with respect to the selected body coordinate system, and therefore, we make a distinction between the kinematics of the coordinate system and the body kinematics.

Fundamental to any presentation of kinematics is an understanding of the rotations in space. This chapter, therefore, is devoted mainly to the development of techniques for describing the orientation of the moving body coordinate system in space. A coordinate system, called hereafter a reference, is a rigid triad vector whose motion can be described by the translation of the origin of the triad and by the rotation about a line deined in the inertial coordinate system. One may then conclude that if the origin of the body reference is fixed with respect to the inertial frame, the only remaining motion is the rotation of the body reference.

In this chapter, approximation methods are used to formulate a finite set of dynamic equations of motion of multibody systems that contain interconnected deformable bodies. As shown in Chapter 3, the dynamic equations of motion of the rigid bodies in the multibody system can be deined in terms of the mass of the body, the inertia tensor, and the generalized forces acting on the body. On the other hand, the dynamic formulation of the system equations of motion of linear structural systems requires the definition of the system mass and stiffness matrices as well as the vector of generalized forces. In this chapter, the dynamic formulation of the equations of motion of deformable bodies that undergo large translational and rotational displacements are developed using the floating frame of reference formulation. It will be shown that the equations of motion of such systems can be written in terms of a set of inertia shape integrals in addition to the mass of the body, the inertia tensor, and the generalized forces that appear in the dynamic formulation of rigid body system equations of motion and the mass and stiffness matrices and the vector of generalized forces that appear in the dynamic equations of linear structural systems. These inertia shape integrals that depend on the assumed displacement field appear in the nonlinear terms that represent the inertia coupling between the reference motion and the elastic deformation of the body.

This chapter provides explanations of some of the fundamental issues addressed in this book. It also provides detailed derivations of some of the important equations presented in previous chapters. The first two sections of this chapter show the detailed derivation of the quadratic velocity centrifugal and Coriolis force vector of Eq. 149 of Chapter 5. The final expression of Eq. 149 of Chapter 5 is obtained using two different approaches; the kinetic energy and the virtual work. It is also shown in Section 3 of this chapter how a general expression of these forces that is applicable to any set of orientation parameters can be obtained. This is the expression used in the generalized Newton–Euler equations presented in Chapter 5 of the book. The generalized centrifugal and Coriolis inertia forces associated with any set of orientation parameters including Euler angles can be obtained from the forces that appear in the Newton–Euler equations using a simple velocity transformation.

Understanding the finite element floating frame of reference formulation presented in Chapter 6 of this book requires a good understanding of the concept of the parallel axis theorem used in rigid body dynamics. The use of the parallel axis theorem is required in rigid body dynamics when the bodies have complex geometric shapes that are characterized by slope discontinuities.

The methods for the nonlinear analysis of physical and mechanical systems developed for use on modern digital computers provide means for accurate analysis of large-scale systems under dynamic loading conditions. These methods are based on the concept of replacing the actual system by an equivalent model made up from discrete bodies having known elastic and inertia properties. The actual systems, in fact, form multibody systems consisting of interconnected rigid and deformable bodies, each of which may undergo large translational and rotational displacements. Examples of physical and mechanical systems that can be modeled as multibody systems are machines, mechanisms, vehicles, robotic manipulators, and space structures. Clearly, these systems consist of a set of interconnected bodies that may be rigid or deformable. Furthermore, the bodies may undergo large relative translational and rotational displacements. The dynamic equations that govern the motion of these systems are highly nonlinear and in most cases cannot be solved analytically in a closed form. One must resort to the numerical solution of the resulting dynamic equations.

The aim of this text, which is based on lectures that I have given during the past several years, is to provide an introduction to the subject of multibody mechanics in a form suitable for senior undergraduate and graduate students. The initial notes for the text were developed for two first-year graduate courses introduced and offered at the University of Illinois at Chicago.

There are two main concerns regarding the use of the classical finite-element formulations in the large deformation and rotation analysis of flexible multibody systems. First, in the classical finite-element literature on beams and plates, infinitesimal rotations are used as nodal coordinates. Such a use of coordinates does not lead to the exact modeling of a simple rigid body motion. Second, lumped mass techniques are used in many finite-element formulations and computer programs to describe the inertia of the deformable bodies. As will be demonstrated in this chapter, such a lumped mass representation of the inertia also does not lead to exact modeling of the equations of motion of the rigid bodies.

In the preceding chapter, a floating frame of reference formulation that uses classical finite-element methodologies is developed. This formulation, in which infinitesimal rotations can be considered as nodal coordinates, can be used only in the large reference displacement and small elastic deformation with respect to the flexible body reference. Using the concept of the intermediate element coordinate system, which is equivalent to the application of the parallel axis theorem used in rigid body dynamics, a nonlinear formulation that leads to exact modeling of the rigid body motion for elements whose coordinates are defined in terms of infinitesimal rotations can be developed.

In the preceding chapter, methods for the kinematic analysis of moving frames of reference were presented. The kinematic analysis presented in the preceding chapter is of a preliminary nature and is fundamental for understanding the dynamic motion of moving rigid bodies or coordinate systems. In this chapter, techniques for developing the dynamic equations of motion of multibody systems consisting of interconnected rigid bodies are introduced. The analysis of multibody systems consisting of deformable bodies that undergo large translational and rotational displacements will be deferred until we discuss in later chapters some concepts related to the body deformation. In the first three sections, a few basic concepts and definitions to be used repeatedly in this book are introduced. In these sections, the important concepts of the system generalized coordinates, holonomic and nonholonomic constraints, degrees of freedom, virtual work, and the system generalized forces are discussed. Although the reader previously may very well have met some, or even all, of these concepts and definitions, they are so fundamental for our purposes that it seems desirable to present them here in some detail. Since the direct application of Newton's second law becomes difficult when large-scale multibody systems are considered, in Section 4, D'Alembert's principle is used to derive Lagrange's equation, which circumvents to some extent some of the difficulties found in applying Newton's second law as demonstrated by the application presented in Section 5.