Anda di halaman 1dari 12

Chapter 1

Introduction
What is a robot? In 1980, the Robot Institute of America (RIA), an industrial trade group, came up with the following denition: A robot is a reprogrammable multifunctional manipulator designed to move material, parts, tools, or specialized devices through variable programmed motions for the performance of a variety of tasks. These days, this denition would be considered too restrictive, as it reects the concentration of the RIA on robot manipulators on an assembly line. Robotics has broadened over the years in many ways: to include mobility platforms, to address the service sector as well as the manufacturing sector, and to incorporate man-machine interactions, not just autonomy, in telerobotic and virtual reality systems. A basic distinction is between mobility and manipulation. A mobile robot can be any form of vehicle, such as a motorized cart, a car, a plane, or a submersible, and in the case of land navigation it can have wheels, tracks, or legs. The main goal of a mobile robot is transport, under guidance of on-board sensors and an intelligent controller.

Figure 1.1: The University of Utahs mobile robot Egor.

c John M. Hollerbach, August 2003

An example of a motorized cart is Egor [13], which is comprised of a LABMATE mobile base, 24 sonar sensors, 8 infrared sensors, a camera and a speaker (Figure 1.1). Egor is shown in a mobile robot competition in a ball-collecting task. Such mobile robot competitions are very popular these days, including soccer competitions and helicopter competitions.

CHAPTER 1. INTRODUCTION

(A)

(B)

Figure 1.2: The University of Utahs (1) CFMMR compliant framed mobile robot, and (2) RAMR2, an articulated climbing robot using suction cups.

More rugged terrain can be handled by the Compliant Framed Modular Mobile Robot CFMMR (Figure 1.2(A)), which has a exible body [1]. Walking and running robots have been built that have one or more legs. The general goal of such locomotion research is to understand biological locomotion and to handle uneven terrain [12]. An example that is a cross between a manipulator and a mobile robot is the climbing robotf RAMR (Recongurable Adaptable Miniature Robot, Figure 1.2(B)), which uses suction cups at either end [11].

These mobile robots all have actuators, sensors, and control systems. Purely passive devices have been built that exhibit walking [10], due to clever kinematics and mass distribution. Such examples emphasize that there can be intelligence in the design, which makes control easy. A manipulator is a mechanical linkage, which may or may not be arm-like, with a gripper or tool to perform some action on the environment. This is the traditional robot that the RIA denition addresses.

An example of an industrial robot at Utah is the PUMA, created by the former Unimation Corp. Other commercial robot vendors include ASEA, Kuka, and Fanuc. Commercial manipulators come in many different shapes and sizes, and vary from very simple pneumatic devices to complicated general-purpose robots.

An example of an arm-like manipulator is the Sarcos Dextrous Arm, a hydraulic manipulator with the same structure of shoulder, elbow, and wrist joints (Figure 1.3(A)). It is primarily employed as a telemanipulator in conjunction with a force-reecting master (Figure 1.6(B)).

The grippers of industrial robots are relatively much simpler than the human hand, and typically are vise-like grippers. Multi-ngered hands such as the human hand have been developed to endow robots with more human-like grasping capabilities.

An example is the Utah/MIT Dextrous Hand (Figure 1.4(A)), which has three ngers and an opposing thumb. The nger joints are actuated by opposing tendons, which are pulled on by electropneumatic actuators [7]. A remotizer routes the tendons to the actuator pack, and mechanically decouples the motion of the hand from the actuator pack (Figure 1.4(B)).

1.1. ROBOTICS APPLICATIONS

(A)

(B)

(C)

(D)

Figure 1.3: (A) Sarcos Dextrous Arm. (B-D) The Morph Hand in various grasping congurations, and demonstrating recongurability to a vice gripper.

(A)

(B)

(C)

Figure 1.4: (A) The Utah/MIT Dextrous Hand. (B) Remotizer, actuator pack and analog joint controller. (C) Tactile sensing system for nger and palm.

An example which is a compromise between the complexity of the human hand and the simplicity of a vise gripper is the Morph Hand for the Sarcos Dextrous Arm (Figure 1.3(B-D)). This hand has three ngers, each of which is a single link. The thumb has two DOFs, while the ngers pinch together. The pads of the ngers can passively rotate to form a vise gripper. Various grasping capabilities for this hand are shown.

In this course, the focus will be on robot manipulators. Basic considerations from manipulators apply to mobile robots and multi-ngered hands.

1.1

Robotics Applications

Historically, there have been two main streams for the development of robotics. One application area is the traditional industrial robot, employed for assembly line tasks such as welding, painting, and materials handling. The other application area is teleoperation for hazardous environments, such as handling nuclear materials. More recently, the domain of application for robotics has expanded greatly.

CHAPTER 1. INTRODUCTION

1.1.1

Industrial Robots

From a commercial standpoint, industrial robotics is still the predominant application area. There are a number of trends in industrial robots. One of the most signicant aspects of robotics is the so-called bin picking problem: how to select objects from a bin of parts so that they are acquired in a known position and orientation. In the past, rather complicated robots, grippers, and vision systems have been applied to this problem, which is still generally unsolved. At the other extreme, vibratory bowl feeders or specially xtured parts feeders have been devised for sorting and feeding specic parts. In between the complete exible robotics solution and the hard automation solution is what has been called minimalist robotics: instead of a complicated and expensive multi-DOF manipulator, very simple mechanisms which might contain simple sensors are employed to push objects so they can be picked up in known position and orientation. Such intelligent xturing systems are based on a very deep understanding of task mechanics [9]. A second trend is xtureless assembly, which goes in the other direction from minimalist robotics. On car assembly lines, manufacturers wish to change rapidly from one model to another, even making different models on the same line. This cannot be done with heavily xtured assembly lines, because these xtures represent hard automation that ties assembly to only the parts for which they have been devised. The idea in xtureless assembly is to have stations in which parts are not precisely presented (say a side panel of a car), but where robot actions are adapted for what is actually there. This requires sophisticated sensing, calibration, and exible control. One reason that such robotics applications have not been developed in the past is that the whole design process for cars was not sufciently computerized.

1.1.2

Service Robots

In recent years, service robots have come increasingly to the fore. Service robotics basically comprise everything that is not industrial robotics, and reects the distinction between the manufacturing and service sectors of the economy. Consequently, service robotics is quite diversied and there are many examples that could be given. Mobile robots have many applications, such as sentry robots and parts transfer. Trucks are being automated for mining applications and storage container transfer. The intelligent highway project is really about the car as a mobile robot. Already passenger jets and large ships are substantially under automatic control. Submersibles are employed for oil exploration and salvage. Rovers on Mars and the moon collect samples and perform analyses. Robots have recently been making a big impact in medicine [15]. Their precision allows surgery to be performed more accurately than human hands are capable of. Examples include hip replacement, craniofacial reconstruction, and neurosurgery. Rehabilitation and the assistance of paralyzed individuals are other emerging applications. Home automation is projected to be a growth area. Cost is a big factor as to whether there will be a robot in your house one day vacuuming the oor, doing dishes, or mowing the grass. One projected need for home robots is elderly care, to reect the ageing of the population. More likely will be the evolution of intelligent appliances, connected to a houses ethernet backbone and under control of the master PC supercomputer. Sophisticated entertainment robots that emulate humans or animals are now found in some major theme parks. A number of such robots have been built by Sarcos, such as certain Disney robots (Figure 1.5(A-B)) and the robot dinosaurs of Jurassic Park the Ride at Universal Studies in Hollywood. Another example is the Ford robot (Figure 1.5(C)), which has been used for new car expositions. Robotic toys are starting to proliferate, ranging from Lego Mindstorms to AIBO, the Sony dog [4].

1.1. ROBOTICS APPLICATIONS

(A)

(B)

(C)

Figure 1.5: (A-B) Disney robots. (C) The Ford robot.

1.1.3

Telerobots

Telerobots are master/slave systems, in which a remote robot (slave) is operated directly by a human through a manual controller (master). The master sends position signals to the slave, which then attempts to follow these position commands. In force-reecting teleoperation, the master is also powered, and force signals are reected back from the slave to the master. The attempt is to give the operator a sense of feel at the remote contact sight, in addition to visual feedback. A basic distinction is between fully automatic control, where a robot acts without human intervention, and fully manual control, where every robot motion is directly human specied. In between there are shades of semi-autonomous control, termed supervisory control [14]. For example, a robot could locally perform some collision avoidance if the directed path would result in a collision. Therefore telerobotics and robotics can be viewed as just lying on the same continuum, rather than representing distinct areas. The main imperative behind telerobotics developments has been hazardous operations. The list is long of examples of such operations: hazardous waste disposal, teleoperated heavy machinery, remote power line maintenance, forestry, pipe inspection, reghting, security, bulk transportation, ocean environments, and space environments [3].

(A)

(B)

(C)

Figure 1.6: (A) The Utah Dextrous Hand Master. (B) The Sarcos Dextrous Arm Master. (C) The Sensuit body tracking system. The masters themselves can be considered robots, which are manipulated by, and in turn manipulate, the

CHAPTER 1. INTRODUCTION

human operator. An example is the Sarcos Dextrous Arm Master (Figure 1.6(B)), which is a force-reecting master that is isomorphic to the Slave. Hydraulic actuators on the Master exert forces on the operator, in response to forces sensed by the slave. Masters may not be actuated, but serve as position sources. Examples include the Utah Dextrous Hand Master (Figure 1.6(A)), a master glove meant for teleoperation of the Utah/MIT Dextrous Hand, and the Sensuit (Figure 1.6(C)), a whole-body goniometer employed to directly control entertainment gures such as the Ford robot. Although technically not robots, such masters are still analyzed by kinematic methods developed for robotics.

1.1.4

Virtual Reality Interfaces

Mechanical interfaces for virtual reality systems are really just masters applied to control and interact with virtual environments, rather than with remote slaves. In fact, virtual reality and teleoperation are very closely related [3]. Position trackers such as joysticks or goniometers such as the Utah Dextrous Hand Master or the Sarcos Sensuit (Figure 1.6(B-C)) are employed to direct an action or an avatar in a virtual environment. For example, in Figure 1.7(A) an operator wearing a Sensuit is controlling a werewolf avatar. Such motion capture is often employed in computer animations, especially in computer games, to give characters lifelike movements. So-called haptic interfaces are just force-reecting masters, whether they be hand controllers, or exoskeletons such as the Sarcos Master (Figure 1.6(B)). Applications for haptic interfaces include simulation and training. For example, the Sarcos Master has been interfaced the Utahs CAD/CAM system, in order to allow a designer to virtually manipulate objects as well as to see them graphically [6]. Shown in Figure 1.7 is an operator directing a virtual hand to trace a disk brake rotor derived from a NURBS surface representation.

(A)

(B)

(C)

Figure 1.7: (A) Sensuit controlling a werewolf avatar. (B) A simulated slave arm interacts with a disk brake rotor derived from a NURBS surface, under guidance of a user employing Sarcos Dextrous Arm Master (C). As haptic interfaces are to arms, locomotion interfaces are to legs. They permit energy expenditure in the virtual environment, to give a more realistic sense of immersion and enhance situational awareness. Several examples have been created by Sarcos, including the Uniport, a powered unicylce, the Biport, two programmable leg platforms, and the Treadport, a treadmill with active mechanical tether (Figure 1.8). The Biport presents uneven terrain, and is comprised of two manipulators which move foot platforms in space. The Treadports mechanical tether measures body position to control treadmill speed and graphics evolution, and also exerts axial force to simulate gravity, unilateral constraints, or inertial forces [2].

1.2. ELEMENTS OF A ROBOT SYSTEM

(A)

(B)

(C)

Figure 1.8: (A) The Uniport, (B) Biport, and (C) Treadport locomotion displays. Concepts from robotics apply to all of these virtual reality interfaces, because they really are robotic systems that interact with humans in particular ways.

1.2

Elements of a Robot System

A robot is comprised of a number of elements, each of which is a worthy focus of attention in its own right.

The mechanical structure comprises the links of a manipulator or the body and wheels of a mobile robot. Links are connected by joints, which dene the degrees of freedom that a robot has. The actuators cause the robot to move. Electric motors, hydraulic actuators, and pneumatic actuators are the common movers of robot joints. Sensors measure robot motion and sense a robots environment. Position and torque sensors may be placed at robot joints, tactile sensors on the ngertips (Figure 1.4(C)), and cameras on a mobile cart. A computer controller reads sensors, sends control signals to actuators, interacts with command sources from an operator, and exhibits various degrees of autonomy.

These varied elements show that robotics is a highly interdisciplinary eld, where contributions from mechanical engineering, electrical engineering, computer science, and even biomechanics intermix. This diversity makes robotics both very interesting and very challenging. The eld of mechatronics describes the general synthesis of design, control, and computation for any electromechanical device. The philosophy of mechatronics is all these aspects come to play in the creation of such devices, and at least some mastery of each aspect is required for a designer to be successful. Robotics can be viewed as a premier example for mechatronics. Similarly, to be successful in robotics, one must have some familiarity with every aspect. It is too much of an undertaking to study all of the elements of a robot from the beginning. Instead, it is normal to start with just the mechanics of robots (kinematics and dynamics) in an introductory course. The mechanical design aspects, the actuation, the sensors, and the real-time controllers can usually be studied in more generic mechanical or electrical engineering courses. Vision sensing is so important that separate courses typically exist for it. High-level intelligence for problem solving and natural language interfaces are subjects for separate courses in articial intelligence.

8
Trajectory planning Position Force x(t) Inverse kinematics

CHAPTER 1. INTRODUCTION

(t)
Position errors Inverse dynamics Force errors

(t)
Torque production

Figure 1.9: A hierarchical sequence of modules representing an advanced controller. Control itself comprises many elements, each of which could be a major area of study. Figure 1.9 shows some of the major components of a control system for a robot manipulator, which is to perform some action such as grasping an object or operating on the environment.

A trajectory plan is formed to move the manipulator from an initial point to a goal point. This plan has to satisfy external constraints, such as avoiding collisions, and internal constraints such as reachability. Most generally, the trajectory is planned in terms of the position and orientation of the end effector, since then it is easiest to comply with the geometry of the environment when operating on it. In addition, external forces may arise during contact, and there may be a plan for endpoint force and torque variables as well.

The inverse kinematics transformation converts end effector position and orientation, and their time derivatives, into joint angles and rates. Potential difculties include singularities, which are manipulator congurations from which certain motions are proscribed, and analytic solvability, because for arbitrary kinematic arrangements a nonlinear search has to be performed. Fortunately, the most common manipulator congurations lend themselves to analytic solutions of the inverse kinematics.

The inverse dynamics transformation derives the joint torques corresponding to desired instantaneous joint positions, velocities, and accelerations. Dynamics of arm movement is rather complicated because movement at one joint generates torques at the other joints, consisting of inertial, Coriolis, and centripetal forces. Efcient recursive algorithms for computing inverse dynamics fortunately exist. Possible complicating factors including exibility of links and joints, and actuator dynamics.

A position controller corrects for errors between actual and desired positions during a trajectory, by generating a corrective torque according to some feedback control law. Such errors may arise from external disturbances or modeling errors. A bewildering assortment of feedback controllers have been proposed, which may or may not incorporate dynamics, adapt to unknown or changing system properties, or formulate control in terms of hand variables. The study of how to formulate robot controllers is in the domain of control theory.

When the manipulator is in contact with an external surface, force as well as position must be controlled. During contact, forces can build up very rapidly with virtually no displacement, and unless

1.3. WHY STUDY ROBOTICS?

the robot or external surface yields, something will break or jam. Either a robot must have a very fast servo response to a force sensor or it must be intrinsically compliant. The control of such interactions is indicated by the force errors module.

1.3

Why Study Robotics?

Robotics is a fascinating discipline, and for some like the author has become a worthy life-long endeavor. On the one hand, there are many exciting and diverse applications, with more being created all the time. On the other hand, robots are interesting objects of themselves: the creation of artifacts with human-like capabilities and reasoning is a compelling and even inevitable goal. Even if one is not interested in robotics as a career, there are many reasons for being familiar with robotics, because this knowledge spills over into other elds.

The concept of pervasive robotics has been emphasized recently. There are robotic aspects to many systems, which themselves are not called robots. This could be kinematic aspects of some linkage, or control aspects of advanced machine tools. The mechatronics combination of disciplines that is a major aspect of robotics is good general training. Robotics majors are often in demand because they have working knowledge of sensors, actuators, electronics, computers, and controllers, and this knowledge is transferable to many other devices. Robotic concepts are important in animation for computer graphics and virtual reality. Whether one controls simulated gures or real robots, the considerations are the same. Concepts developed in the modeling of object interaction for assembly planning apply directly to modeling moving objects in virtual environments.

Robotic concepts can often aid in the understanding of corresponding biological capabilities. Obviously, the human arm accomplishes many tasks that we seek to develop for robot arms. Concepts developed for robot arms may correspondingly give us more insight into problems confronted by human arms. This is not to say that robotic solutions are also candidates as biological solutions, only that by attempting to duplicate human function we can better appreciate it [5].

1.4

Goals of This Course

This course teaches the fundamentals of robotics: kinematics, dynamics, and control. The approach is to provide a strong underpinning, rather than a cookbook approach, in order that the student may more readily apply concepts to new situations. The material has been chosen to reect actual robotics practice. Introductory robotics texts often stop short of presenting this practice, which misleads students into thinking they have the appropriate tools for robotics applications. Consequently, this course delves into greater depth than most introductory robotics courses. Kinematics is the traditional starting point for robotics courses, and is by far the most important and deepest topic. Kinematics has broad application to other elds such as graphics and computer animation. Kinematics requires the least prerequisites, basically just linear algebra and differential equations. On the other hand, the linear algebra background is exercised thoroughly. Students often go through linear algebra without rmly having grasped all concepts, and robot kinematics is a good application area to engrain this knowledge.

10

CHAPTER 1. INTRODUCTION

Just as with kinematics, there has developed in robotics a preferred way of representing manipulator dynamics. Many students will not have had much exposure to concepts in dynamics, or if they had it was with a different approach. This course teaches dynamics from scratch, assuming only the most rudimentary knowledge of mechanics from freshman physics and volume integrals from calculus. Yet the development of concepts from scratch will result in a more advanced knowledge of dynamics than found in introductory robotics texts, which just state relationships without derivation or explanation. Robot control is an application of control theory, which is a huge and sophisticated academic topic on its own. Again, many students will not have had exposure to an introductory control theory course. Our approach will be to just provide the basic approaches to robot control, and leave it to the interested student to pursue an interest in control with specic control courses. The course utilizes Matlab in some of its homework assignments. Matlab has become a standard numerical package, and students will nd it useful in many other aspects of their careers.

Bibliography
[1] Albiston, B.W., and Minor, M.A., Curvature based point stabilization for compliant framed wheeled modular mobile robots, in Proc. IEEE Intl. Conf. Robotics and Automation, Taiwan, 2003. [2] Christensen, R., Hollerbach, J.M., Xu, Y., and Meek, S., Inertial force feedback for the Treadport locomotion interface, Presence: Teleoperators and Virtual Environments, 9, 2000, pp. 1-14. [3] Durlach, N.I., and Mavor, A.S., eds., Virtual Reality: Scientic and Technological Challenges. Washington, D.C.: National Academy Press, 1994. [4] Fujita, M., AIBO: towards the era of digital creatures, in: Robotics Research: the Ninth International Symposium, J. Hollerbach and D. Koditschek (Eds), Springer-Verlag, London, 2000, pp. 315-320. [5] Hildreth, Ellen C., and Hollerbach, J.M., Articial Intelligence: computational approach to vision and motor control, in: Handbook of Physiology, Section 1: The Nervous System, Volume V: Higher Functions of the Brain, Part II, F. Plum, ed., American Physiological Society, Bethesda, Maryland, 1987, pp. 605-642. [6] Hollerbach, J.M., Cohen, E., Thompson, W., Freier, R., Johnson, D., Nahvi, A., Nelson, D., Thompson, T., and Jacobsen, S., Haptic interfacing for virtual prototyping of mechanical CAD designs, ASME Design for Manufacturing Symposium, Sacramento, Sept. 14-17, 1997. [7] Jacobsen, S.C., Iversen, E.K., Knutti, D.F., Johnson, R.T., and Biggers, K.B., Design of the Utah/MIT Dextrous Hand, in Proc. IEEE Int. Conf. Robotics and Automation, San Francisco, pp. 1520-1532, April 7-10, 1986. [8] Jacobsen, S.C., Smith, F.M., Iversen, E.K., and Backman, D.K., High performance, high dexterity, force reective teleoperator, in Proc. 38th Conf. Remote Systems Technology, Washington, DC, pp. 180-185, Nov., 1990. [9] Mason, M.T., and Salisbury, J.K., Robot Hands and the Mechanics of Manipulation. Cambridge, MA: MIT Press, 1985. [10] McGeer, T., Passive dynamic walking, Int. J. Robotics Research, vol. 9, no. 2, pp. 62-82, 1990. [11] Minor, Mark A., and Mukherjee, Ranjan, Under-actuated kinematic structures for miniature climbing robots, J. Mechanical Design, vol. 125, pp. 281-291, 2003. [12] Raibert, M.H., Legged Robots That Balance. Cambridge, MA: MIT Press, 1986. [13] Schenkat, L., Veigel, L., Henderson, T.C., EGOR: Design, Development, Implementation An Entry in the 1994 AAAI Robot Competition, UUCS-94-034, University of Utah, Dec., 1994. 11

12

BIBLIOGRAPHY

[14] Sheridan, T.B., Telerobotics, Automation, and Human Supervisory Control. Cambridge, MA: MIT Press, 1992. [15] Taylor, R.H., Lavallee, S., Burdea, G., and Mosges, R., Computer Integrated Surgery: Technology and Clinical Applications. MIT Press, 1996.