National Academies Press: OpenBook
« Previous: Simulation in Manufacturing
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

ROBOTICS

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
This page in the original is blank.
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

A Brief History of Robotics

Kenneth Y. Goldberg

Department of Industrial Engineering and Operations Research University of California Berkeley, California

Almost all cultures have myths about artificial creatures: the Homunculus, Golem, Sorcerer's Apprentice, and, of course, Frankenstein. One theme that runs through these myths is that things generally turn out rather badly for the mortals who attempt to play creator.

In 1923 the playwright Karel Capek coined the word "robot"—"worker" in Czech. Industrialization was in full swing; Capek's extrapolation of this trend, combined with the development of the electronic computer in the 1940s, fired the imagination of many science fiction writers. Robots were the focus of Asimov's I Robot series and were featured in many popular films (2001) and television shows (Lost in Space).

What is a robot? The definition is surprisingly controversial, even among roboticists. At one extreme are humanoids, friendly or unfriendly, with anthropomorphic features. At the other extreme are the repetition-loving mechanical arms of industrial automation. The former are highly flexible, the latter highly efficient. Roboticists such as Whitney (1986) suggest that there is an inherent design tradeoff between flexibility and efficiency: a humanoid household robot would not be nearly as efficient at the standard home dishwashing machine. Neither is it obvious that anthropomorphism is necessary: machines with flapping wings were far less successful at flying than fixed-wing aircraft.

The International Standards Organization defines an industrial robot as "an automatically controlled, reprogrammable, multipurpose, manipulator with three or more axes" (ISO 8373), a reasonable definition that excludes dishwashers and most talking dolls. The first U.S. patent for a robot that falls under this definition was granted to George Devol in 1956. In the 1960s, research and develop-

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

ment on numerical control (NC) machines for milling and lathing grew into a large commercial industry.

In the early 1980s, around the time the film Star Wars came out, robots became America's media darlings. They were suddenly friendly and desirable. Intoxicating predictions were made about robots leading to the end of human labor. Many companies invested in robots but faced the challenge of making them work reliably. Unrealistic expectations led to disillusionment and the failure of many robot companies at the end of that decade.

Today, there are almost 700,000 robots at work in industry. Approximately 80,000 robots were sold throughout the world in 1996, accounting for gross sales of over $5 billion. Almost half of those robots were installed in Japan, about 10,000 were installed in the United States and Germany, respectively, and the remaining 20,000 were installed in Korea, Italy, France, and other countries. By far the largest application areas are welding and painting, followed by machining and assembly. The largest customers are the automotive industry, followed by electronics, food, and pharmaceuticals (United Nations/Economic Commission for Europe and International Federation of Robotics, 1997).

Throughout this turbulence robotics continues to be an active and thriving area of research. Some of its subareas are kinematics (the study of positions and velocities), dynamics (the study of forces), and motion planning (how to get an object from here to there while avoiding obstacles). The so-called "piano movers" problem was solved in the 1980s when a breakthrough showed that it could be reduced to a well-known problem of deciding the truth value of algebraic sentences (Latombe, 1991). Other areas include grasping, locomotion, actuator design (the direct-drive robot arm was another breakthrough in the 1980s), and sensor design (a reliable tactile sensor is still being pursued).

Robotics is highly interdisciplinary, including specialists from fields such as mechanical engineering, computer science, electrical engineering, and industrial engineering. Research is sponsored in this country by the National Science Foundation, the Defense Advanced Research Projects Agency, the U.S. Department of Energy, the National Aeronautics and Space Administration, and industry. Similar sponsors are found for research in Japan, Australia, and Europe. Almost all universities have research groups working in robotics. The largest international research organization is the Institute of Electrical and Electronics Engineers Society of Robotics and Automation, founded in 1984.

There are many frontiers of robotics. Theories from areas such as nonlinear control, Lie algebra, computational geometry, and computational algebra are being applied to such topics as medical and surgical robots, microscale robots, Internet robots, modular robots, and robot toys for education and entertainment. It is impossible to cover all of these frontiers in one session. For the Frontiers symposium I worked closely with Susan Corwin of Intel and Rob Howe of Harvard to select the following four roboticists to represent a cross section of the best new work in our field.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

References

Latombe, J.-C. 1991. Robot Motion Planning. Boston: Kluver Academic Publishers.


United Nations/Economic Commission for Europe and International Federation of Robotics. 1997. World Industrial Robots 1997—Statistics, Analysis, and Forecasts to 2000. New York: United Nations.


Whitney, D. E. 1986. Real robots don't need jigs. Proceedings of the IEEE International Conference on Robotics and Automation, Washington, D.C.: IEEE Computer Society Press.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Algorithms in Robotics: The Motion Planning Perspective

Lydia E. Kavraki

Department of Computer Science Rice University Houston, Texas

The development of robotics depends not only on the advancement of robotic technology and the building of novel robotic systems but also on improved understanding and design of robot algorithms. Like computer algorithms, robot algorithms are abstract descriptions of processes in the physical world whose development and analysis are, to some extent, independent of a particular implementation technology. But robot algorithms also differ in significant ways from computer algorithms. While the latter have full control over the data to which they apply, robot algorithms deal with physical objects that they attempt to control despite the fact that these objects are subject to the independent and imperfectly modeled laws of nature. This leads robot algorithms to blend in a unique way basic control issues (controllability and observability) and computational issues (calculability and complexity). They pose in turn fundamental questions that require a new set of analytical concepts and tools for the evaluation of their performance (Latombe, 1994).

Among robot algorithms, those that deal with the planning of continuous motion are central to robotics. Robots accomplish tasks by moving objects (including themselves) in the real world, and they need to reason about the continuous geometry and physics of their environment. The generation of motion clearly poses both control and computational problems. A number of key results have influenced recent work. In particular, the computational complexity of several motion planning problems was established by using theoretical computer science methodology (Reif, 1979). The concept of the configuration space permitted discussion of motion in a unified framework (Lozano-Pérez, 1983). Under this framework, the problem of reasoning about a robot in its original environment is transformed to the problem of reasoning about a point in a new space

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

that is called the configuration space and whose dimension equals the number of degrees of freedom of the robot. Critically based decomposition techniques first addressed the problem of computing with continuous geometric data by partitioning the space into regions such that some pertinent property remains invariant over each region (Schwartz and Sharir, 1983). Moving away from purely geometric issues, the subtle relations between state reachability and state recognizability were studied (Lozano-Pérez et al., 1984), and the role of task mechanics in the generation of motion was emphasized (Mason, 1986). As with robot algorithms in general, the amount of physical resources required (number of hands, beacons, etc.) defines the physical complexity of the algorithm. Understanding the intrinsic physical complexity of motion tasks has aided the design of reliable algorithms that induce minimal engineering costs (Canny and Goldberg, 1994; Donald, 1993).

The following focuses on the computational issues of motion planning by examining the basic problem of planning a continuous geometric path between two free configurations of a robot. The geometry of the robot, the kinematics of the robot, and the geometry of the environment are completely known. The path planning problem is a central research area in robotics and several variants of the problem have been explored (Latombe, 1991). These include parts manipulation, assembly sequencing, multirobot coordination, on-line execution, controllability, planning with incomplete knowledge or uncertainty in control and sensing, and others. Currently, the application of planning techniques extends well beyond the realm of traditional robotics to applications in medicine (robot-assisted surgery), computational biology and computational chemistry (computer-assisted pharmaceutical drug design), and graphics animation (digital actors).

Complete algorithms are known for the path planning problem. However, the fastest among them is exponential in the number of degrees of freedom of the robot (Canny, 1988), and it is unrealistic, from a computational point of view, to apply such an algorithm to robots with more than two or three degrees of freedom. Nevertheless, remarkable progress has been made in practical path planning over the past few years. Planners with weaker performance guarantees (probabilistically complete planners) have been developed for robots with many degrees of freedom with increasingly improved performance. An important milestone was the introduction of randomization techniques for solving high-dimensional problems (Barraquand and Latombe, 1991). Recently, a unified randomized framework has dealt successfully with 5 to 25 degrees-of-freedom robots (Kavraki and Latombe, 1994). The analysis of this framework, called the probabilistic roadmap approach to path planning, demonstrates how the geometric properties of the underlying space (other than the number of vertices, edges, faces, etc., of the objects in the environment) can be exploited to yield a planner with performance guarantees.

The probabilistic roadmap approach proceeds as follows. During a preprocessing phase a roadmap is constructed. This is done by first generating

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

random free configurations of the robot and then connecting these configurations using a simple, but very fast, motion planner. The roadmap thus formed in the free configuration space of the robot is stored as an undirected graph. The configurations are the nodes of the graph, and the paths computed by the local planner are the graph edges. A subsequent processing of the graph attempts to increase its connectivity again by using probabilistic techniques. Following the preprocessing phase, multiple path planning queries can be answered. A query asks for a path between two free configurations of the robot. To answer a query, the planner first attempts to find a path from the start and goal configurations to two nodes of the roadmap. Next, a graph search is performed to find a sequence of edges connecting these nodes in the roadmap. Concatenation of the successive path segments transforms the sequence found into a feasible path for the robot.

Importantly, the performance of the method can be related to basic geometric properties of the space in which the problem is solved (Kavraki et al., 1995). The evaluation shows that in spaces that are ε-good (in these spaces each point "sees" an ε fraction of the whole free space), a roadmap with polynomial in ε number of nodes captures the connectivity of the free space with high probability. This allows path planning queries to be answered efficiently and reliably once such a roadmap has been constructed. The analysis explains the good experimental performance of the planner in realistic settings and introduces a new methodology for evaluating the performance of planning algorithms.

Recent work is extending the probabilistic framework above to planning for multiple robots, planning for nonholonomic robots (robots with no-slip constraints, such as carllike robots), planning in environments with moving objects, and planning for flexible objects. Introducing physical properties in geometric settings (as done in the case of flexible objects; Kavraki et al., 1998) gives rise to even more difficult discretization and sampling issues. Furthermore, the probabilistic framework is used to plan paths for drug molecules to their docking sites in proteins (A. Singh, J.-C. Latombe, and D. Brutlag, Stanford University, personal communication, 1998). Interestingly, work in robotics is finding many applications in the area of computer-assisted pharmaceutical drug design (Kavraki, 1997).

Understanding of the intrinsic geometric properties of motion tasks and the development of efficient techniques that rely on these properties is still at an early stage. It is now becoming necessary to study these issues in depth and develop a computational framework that can directly relate the difficulty of tasks to their geometric and physical characteristics. The goal is to allow, with minimal effort, the tailoring of robot planning algorithms to a variety of tasks-from specialized operations in extremely constrained environments to everyday life activities.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

References

Barraquand, J., and J.-C. Latombe. 1991. Robot motion planning: A distributed representation approach. International Journal of Robotics Research 10(6):628-649.


Canny, J. 1988. The Complexity of Robot Motion Planning. Cambridge, Mass.: MIT Press.

Canny, J., and K. Y. Goldberg. 1994. RISC for industrial robotics: Recent results and open problems. Pp. 1951-1958 in Proceedings of the IEEE International Conference on Robotics and Automation. Washington, D.C.: IEEE Computer Society Press.


Donald, B. 1993. Information invariants in robotics. Pp. 276-283 in Proceedings of the IEEE International Conference on Robotics and Automation. Washington, D.C.: IEEE Computer Society Press.


Kavraki, L. 1997. Geometry and the discovery of new ligands. Pp. 435-448 in Algorithms for Robotic Motion and Manipulation, J.-P. Laumond and M. Overmars, eds. Wellesley, Mass.: AK Peters.

Kavraki, L., and J.-C. Latombe. 1994. Randomized preprocessing of configuration space for fast path planning. Pp. 2138-2145 in Proceedings of the IEEE International Conference on Robotics and Automation. Washington, D.C.: IEEE Computer Society Press.

Kavraki, L., J.-C. Latombe, R. Motwani, and P. Raghavan. 1995. Randomized query processing in robot motion planning. Pp. 353-362 in Proceedings of the 27th Annual ACM Symposium on the Theory of Computing. New York, N.Y.: Association for Computing Machinery.

Kavraki, L., F. Lamiraux, and C. Holleman. 1998. Towards planning for elastic objects. Pp. 313-325 in Robotics: The Algorithmic Perspective, P. Agarwal, L. Kavraki, and M. Mason, eds. Wellesley, Mass.: A K Peters.


Latombe, J.-C. 1991. Robot Motion Planning. Boston, Mass.: Kluwer.

Latombe, J.-C. 1994. Robot algorithms. Pp. 1-8 in Algorithmic Foundations of Robotics , K. Y. Goldberg, D. Halperin, J.-C. Latombe, and R. H. Wilson, eds. Wellesley, Mass.: A K Peters.

Lozano-Pérez, T. 1983. Spatial planning: A configuration space approach. IEEE Transactions on Computers C-32(2):108-120.

Lozano-Pérez, T., M. Mason, and R. Taylor. 1984. Automatic synthesis of fine-motion strategies for robots. International Journal of Robotics Research 3(1):3-24.


Mason, M. 1986. Mechanics and planning of manipulator pushing operations. International Journal of Robotics Research 5(3):53-71.


Reif, J. 1979. Complexity of the mover's problem and generalizations. Pp. 421-427 in Proceedings of the 20th Annual Symposium on Foundations of Computer Science. New York, N.Y.: Institute of Electrical and Electronics Engineers.


Schwartz, J., and M. Sharir. 1983. On the "piano movers" problem. II. General techniques for computing topological properties of real algebraic manifolds. Advances in Applied Mathematics 4(3):298-351.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Mechanics, Control, and Applications of Biomimetic Robotic Locomotion

Joel W. Burdick

Department of Mechanical Engineering California Institute of Technology Pasadena, California

Introduction

Since mobility can be an essential requirement for the operation of many autonomous systems, robotic locomotion has been actively studied for over three decades. Most mobile robots use wheels, since they provide the simplest means for mobility. However, some terrains are inaccessible to wheeled vehicles, and wheels are undesirable for a number of applications. Biomimetic locomotion refers to the movement of robotic mechanisms in ways that are analogous to the patterns of movement found in nature. Biomimetic robotic locomotors do not rely on wheels, tracks, jets, thrusters, or propellers for their propulsion.

Despite more than three decades of research effort, biomimetic robotic locomotors have largely remained laboratory curiosities. However, promising applications of biomimetic mobility still motivate research in this area. The potential for robots to assist the elderly in their homes has recently motivated the Honda Motor Co. (1998) to invest heavily in bipedal walking research, since anthropomorphic robots can adapt better to preexisting environments that were designed for humans. In the 1980s, the Defense Advanced Research Projects Agency provided significant support for Ohio State University's Adaptive Suspension Vehicle, a six-legged, 7,000-pound locomotor that was intended to support combat operations in complex terrain (Song and Waldron, 1989). Nonlegged locomotory machines might find significant applications as well. For example, efforts are under way at Caltech to develop a "snakelike" robotic endoscope that would enable minimally invasive access to the human small bowel system, which is currently inaccessible by conventional endoscopes. Snakelike robots have also been investigated for use in urban search-and-rescue operations following earthquakes or other natural disasters. Biomimetic fluid propulsion based on changes

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

in the mechanism's shape offers an alternative to traditional underwater vehicle propulsion based on propellers and control surfaces. Such fluid propulsion may be very maneuverable at many size scales and is free from the motor noise, vibrations, and propeller cavitation associated with propellers.

Current Status

Quasi-static legged locomotion (where the locomotor's center of mass is always supported by at least three legs in ground contact) has been the most extensively studied type of biomimetic locomotion, and many four- and six-legged robots have been successfully demonstrated (Song and Waldron, 1989). Beginning with Raibert (1986), legged hopping robots have received considerable experimental and analytical attention (M'Closkey and Burdick, 1993). Bipedal walking and running also has been an active area of study (McGeer, 1990; Honda Motor Co., 1998). "Snakelike" robots can potentially enter environments that are inaccessible to legged or wheeled vehicles. Significant work in snakelike locomotion was initiated in the 1970s by Hirose and Umetani (1976) and more recently by Chirikjian and Burdick (1995). Realistic efforts to develop fishlike robots have emerged only recently (Barrett, 1996; Kelly et al., 1998).

Research Needs and Objectives

There are many issues that limit widespread deployment of biomimetic locomotors, including limitations in actuation technology, onboard power-carrying capability, and sensing. While all of these issues merit serious attention, this paper briefly discusses the associated limitations in theory. In an attempt to derive useful results for specific examples, prior biomimetic locomotion studies have generally focused on a particular robot morphology (such as a biped or quadruped). Unfortunately, results derived for one morphology typically do not extend to other morphologies. To enable future widespread deployment of cheap and robust robotic locomotion platforms, we must ultimately seek a more unifying and comprehensive framework for biomimetic robotic locomotion engineering. This framework should have the following properties:

  • the analysis, design, and control methodologies can be uniformly applied to a broad class of locomotory problems;
  • significant aspects of the framework can be encoded in automated software tools; and
  • the underlying methods are sufficiently rigorous to predict and enable robust system performance.

Realization of such a framework would enable more widespread deployment of effective biomimetic locomotors. One strategy being pursued at Caltech

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

to realize this framework is to (1) establish general forms for the equations of motion of locomotion systems, (2) develop a control theory for this class of nonlinear equations, (3) abstract motion planning and feedback control algorithms from the control theory, and (4) develop paradigms (perhaps rules of thumb) for designing systems for specific applications. Caltech's research program, whose underlying concepts are briefly sketched in the next section, has reached a good understanding of the underlying mechanics principles. The associated control theory and algorithms are currently in development.

Principles of Biomimetic Propulsion

Biomimetic propulsion is typically generated by a coupling of periodic mechanism deformations to external constraints (i.e., mechanical interactions with the environment). The forces generated by these constraint interactions (e.g., pushing, rolling, sliding) induce net robot movement. The creeping, sidewinding, and undulatory gaits of snakes rely on no-slip, or nonholonomic, constraints. Slug and snail movement depends on the viscous fluid constraint of slime trails, while amoebae and paramecia move via a constraint between their surfaces and the surrounding fluid. Fish use a variety of fluid mechanical constraint principles. Surprisingly, common principles underlie the mechanics and control of these seemingly different systems (Ostrowski and Burdick, 1998).

The language of geometric mechanics has proven to be a useful way to precisely phrase these intuitive notions. Two simple concepts motivated by geometric thinking have proven useful in developing a comprehensive basis for the mechanics and control of biomimetic locomotion. The first key observation is that it is always possible to divide a locomoting robot's configuration variables into two classes. The first class of variables describes the position of the robot—that is, the displacement of a robot fixed coordinate frame with respect to a fixed reference frame. The set of frame displacements is SE(m ), m ≤ 3, or one of its subgroups—that is, a Lie group. The second set of variables defines the mechanism's internal configuration or shape. The set of all possible shapes (the "shape space") is a manifold, M. The Lie group, G, together with the shape space, M, form the total configuration space of the system, denoted by Q = G × M. The configuration space of both terrestrial and aquatic biomimetic locomotors is a trivial principal fiber bundle.

The importance of the principal fiber bundle structure of the configuration space of locomoting systems is related to the following facts. The shape and position variables are coupled by the constraints acting on the robot. By making changes in the shape variables, it is possible to effect changes in the position variables through the constraints. A central goal of locomotion analysis is the systematic derivation of an expression that answers the question: If I wiggle the body, how far does the mechanism locomote? Formally, this all-important relationship between shape changes and position changes can be described via a

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

connection, which is an intrinsic geometric feature of principal fiber bundles. Recent Caltech efforts have shown that there is a systematic way to derive the connection for a very large class of locomotion problems. The connection provides not only a unified way of thinking about mechanics but also one about motion planning and control (Radford and Burdick, 1998).

The second key idea is the use of symmetries in locomotion analysis. Because a locomotor is a mechanical system, we can assume there exists a Lagrangian and a set of constraint equations that describe the interaction principle underlying the given locomotion scheme. A symmetry corresponds to a group of transformations that leave the Lagrangian (and possibly the constraints) invariant. In the absence of constraints, these symmetries correspond to the well-known principles of momentum conservation. Unfortunately, conservation laws are not necessarily preserved in the presence of most constraints, which are essential to locomotion. However, it is possible to extend classical theory (see Bloch et al., 1996, for the case of nonholonomic constraints) to develop a generalized momentum equation that describes the evolution of the momenta due to the interaction constraints. The mixture of symmetry and interaction constraints can give rise to the ability to increase or control momentum via the action of internal forces. This is an extremely important effect in generating biomimetic locomotion. The aforementioned connection and the use of invariance principles (and their associated momenta equations) yield a comprehensive framework for the fundamental analysis of biomimetic locomotion mechanics and control.

References

Barrett, D. S. 1996. Propulsive Efficiency of a Flexible Hull Undersea Vehicle. Ph.D. dissertation. Massachusetts Institute of Technology, Cambridge.

Bloch, M., P. S. Krishnaprasad, J. E. Marsden, and R. M. Murray. 1996. Nonholonomic mechanical systems with symmetry. Archive for Rational Mechanics and Analysis 136(1):21-99.


Chirikjian, G. S., and J. W. Burdick. 1995. The kinematics of hyper-redundant robot locomotion. IEEE Transactions on Robotics and Automation 11(6):781-793.


Hirose, S., and Y. Umetani. 1976. Kinematic control of active cord mechanism with tactile sensors. Pp. 241-252 in Proceedings of the Second International CISM-IFT Symposium on Theory and Practice of Robots and Manipulators, A. Morecki and K. Kñedzior. eds. New York: Elsevier Scientific Publishers.

Honda Motor Co., Ltd. 1998. Humanoid robot. [Online]. Available: http://www.honda.co.jp/home/hpr/e_news/robot/index.html."


Kelly, S., R. Mason, C. Anhalt, R. Murray, and J. Burdick. 1998. Modeling and experimental investigation of carangiform locomotion for control. Paper presented at the American Control Conference, Philadelphia, Pa., June 24-26, 1998.


McGeer, T. 1990. Passive dynamic walking. International Journal of Robotics Research 9(2):62-82.

M'Closkey, R. T., and J. W. Burdick. 1993. On the periodic motions of a hopping robot with vertical and forward motion. International Journal of Robotics Research 12(3): 197-218.


Ostrowski, J. P., and J. W. Burdick. 1998. The mechanics of undulatory robotic locomotion. International Journal of Robotics Research 17(7):683-701.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Radford, J., and J. Burdick. 1998. Local motion planning for nonholonomic control systems evolving on principal bundles. Paper presented at the International Symposium on Mathematical Theory of Networks and Systems (MTNS '98), Padova, Italy, July 6-11, 1998.

Raibert, M. H. 1986. Legged Robots that Balance. Cambridge, Mass.: MIT Press.


Song, S. M., and K. J. Waldron. 1989. Machines that Walk: The Adaptive Suspension Vehicle. Cambridge, Mass.: MIT Press.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Robotic Perception for Autonomous Navigation of Mars Rovers

Larry H. Matthies

Jet Propulsion Laboratory Pasadena, California

The U.S. government has been working on autonomous navigation of robotic vehicles since the 1960s, when the Jet Propulsion Laboratory (JPL) developed a prototype lunar rover for the Surveyor program. The earliest U.S. and Russian rovers were essentially teleoperated, which was acceptable for the few seconds of communications delay between earth and the moon. In the 1970s, JPL began working on rovers for Mars, where communication latency of up to 40 minutes, owing to roundtrip light time, required a higher degree of autonomy on the rover if exploration was to be efficient (O'Handley, 1973). Key technical barriers were relatively poor three-dimensional (3-D) sensing of the environment, lack of accurate means to keep track of the position of the rover as it traveled, and the low performance of onboard computers. This paper surveys key progress in sensors, algorithms, and processors that has alleviated these specific barriers, thereby enabling practical autonomous navigation.

Background

Research on robotic vehicles for terrestrial applications accelerated in the late 1970s and throughout the 1980s. The Autonomous Land Vehicle (ALV) Program, funded by the Defense Advanced Research Projects Agency (DARPA) from 1984 to 1988, developed a vehicle the size of a large bread truck that used a scanning laser range finder to locate and avoid obstacles while driving cross-country at speeds on the order of 2 mph (Olin and Tseng, 1991). Progress in this area was retarded by the size, cost, and power consumption of the laser rangefinder and by the size and cost of the onboard computers. Meanwhile, in the more structured environment of autonomous lane-following on highways, researchers

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

in Germany were able to achieve driving speeds of up to 60 miles an hour for a Mercedes van that used onboard cameras and computers to track lane markings (Dickmanns and Mysliwetz, 1992). This application was helped by the strong geometric constraints available from knowledge of highway design and by the relatively low amount of computation required to track lane markings in imagery. Nevertheless, this effort made a significant contribution to the field by demonstrating effective application of system dynamics models and Kalman filtering in a computer vision system for autonomous navigation.

Real-Time 3-D Perception For Mars Rovers

A key breakthrough in 3-D perception for autonomous cross-country navigation came in 1990, when JPL developed efficient and reliable algorithms for estimating elevation maps in real time from stereo image pairs using compact commercial processors onboard a prototype Mars rover (Matthies, 1992). Three-dimensional sensing with stereo cameras has advantages over laser scanners for Mars rover applications because it is easier to make the sensor mechanically robust enough to survive the rigors of launch and landing. This stereo vision approach computes range by triangulation, given matching features in two images. The prevailing approach to stereo matching in the computer vision community at that time, popularized by David Marr (1982) at the Massachusetts Institute of Technology, was to first extract edges from both images and then match the edges found in the left and right images. The edge detection process reduces the amount of information to be processed in the matching stage, which is what made this approach appear attractive; however, by definition it produces "sparse" range data because it measures range only where prominent edges are found in the imagery. For Mars rover applications this would mean that range data would only be found around the outlines of high-contrast rocks. However, to do effective traverse planning, it was preferable to sample the elevation of the terrain more densely, so as to be guaranteed of finding small obstacles that might trap a wheel.

The heart of JPL's innovation was to develop an efficient stereo matching algorithm based on area correlation, which was able to produce reliable range measurements at almost every pixel in the image for applications like Mars rovers. Most components of this algorithm had been developed previously for different problems: hence, the key contribution was to recognize how to put them together to produce a fast and reliable solution to this problem. The first step in this algorithm is a process called rectification, which resamples both images in such a way that corresponding image features lie on corresponding scanlines in the two images. This reduces the search for matching features to a I-D search along the corresponding scanlines. The second step applies a band-pass filter to each image to compensate for overall differences in brightness between images from the two cameras. The third step finds corresponding fea-

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

tures by evaluating a least squares similarity criterion for a small image patch from the left image at several trial match positions along the scanline in the right image. This evaluation is performed for each pixel in the left image to produce a range estimate at each pixel; the large number of correlations to perform makes this the bottleneck step in the whole algorithm. In straightforward implementations the number of arithmetic operations needed to evaluate the least squares criterion is 3N, not including addressing operations, where N is the number of pixels in the image patch used for matching; N is typically around 50 (i.e., a 7×7 patch). However, there exists an incremental technique that maintains intermediate results to reduce the cost to six operations per trial match position, independent of the size of the image patch used for matching. This was key to making the entire process practical for real-time implementation onboard a robotic vehicle.

The first incarnation of a stereo vision system using this algorithm required nearly 10 processor boards occupying about I cubic foot of space and using 100 to 200 watts of power. This system produced about 1,000 range measurements per second; this enabled the JPL Mars rover prototype "Robby" to drive autonomously over a 100-meter cross-country course in four hours in September 1990 (Matthies, 1992)—a significant "first" for a robotic vehicle.

Further Development of Real-Time 3-D Perception

This technology was picked up by DARPA and the U.S. Army for use in research programs aimed at developing unmanned ground vehicles for military reconnaissance applications (Mettala, 1992). By 1996 the speed of stereo vision systems had increased to about 30,000 range measurements per second, with faster computers that occupied slightly less space. This enabled semiautonomous High Mobility Multipurpose Wheeled Vehicles (HMMWVs) to execute rudimentary reconnaissance missions covering about 2 miles of open terrain at speeds up to 5 mph (Matthies et al., 1996).

Since 1996, the advent of general-purpose microprocessors with limited vector processing capability has enabled substantial speed improvements and size reductions for stereo vision systems. The MMX feature in the Intel Pentium is the best-known example of such a capability; it allows up to four 16-bit integer operations to be performed in parallel. Stereo algorithms implemented on these processors can now perform about 700,000 range measurements per second (Konolige, 1998). Lower-performance systems based on the same algorithm but different CPUs have been built on circuit cards that fit in the palm of the hand, including the CPU and both cameras. This highlights another recent development that is enabling compact, low-cost computer vision systems for robotic vehicles: the introduction of low-power complementary metal-oxide silicon (CMOS) imagers with clocking, control, and analog-to-digital conversion functions fully integrated onchip. Forthcoming advances in vector processing for

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

general-purpose CPUs will accelerate these trends; for example, Motorola has announced a vector-processing extension to the PowerPC architecture that will allow 16 byte-wide arithmetic operations to proceed in parallel. Within a year this will enable stereo vision systems to produce on the order of 2 million range measurements per second with a single microprocessor—that is, 256×256 range measurements per frame at full video rate (30 frames per second). Compared to 1990, this is a speed increase of three orders of magnitude with a simultaneous reduction in size and power dissipation of one order of magnitude.

These advances are enabling a suite of new applications of robotic vehicles. In 1999 a robotic vehicle carrying stereo cameras is to enter Chernobyl to attempt to create a 3-D model of the interior to facilitate further cleanup efforts. JPL's stereo algorithm will help enable the next U.S. Mars rover, currently scheduled for launch in 2003, to explore several kilometers, in comparison with the 100 meters or so covered by the Sojourner rover in the summer of 1997. DARPA and the U.S. Army are also continuing to use this technology for further development of military robotic vehicles. For example, the Demo III program managed by the Army Research Laboratory aims to enable autonomous cross-country navigation at 20 mph by 2001, for a robotic vehicle the size of a large desk. The DARPA Tactical Mobile Robot Program is currently funding development of robotic vehicles the size of a large briefcase for reconnaissance applications in urban warfare. Both of these programs will employ stereo vision among their sensor suites for autonomous navigation and will depend on the aforementioned advances in algorithms, low-power CMOS imagers, and high-performance embedded CPUs to provide the increased speed and smaller size required.

Limitations and Approaches to Solutions

These advances have produced a viable solution to real-time 3-D perception for robotic vehicles operating during the day in barren or semiarid terrain. Limitations that arise as we push for broader applicability include the following:

  • For military applications, operability at night is essential. It appears that stereo vision with thermal infrared imagery works quite well, although thermal cameras are currently very expensive. Two-axis scanning laser rangefinders work well at night but are also still large and expensive.
  • Stereo vision fails in textureless environments, such as painted walls in indoor mobile robot applications. This can be solved by adding low-cost, low-resolution active sensors (e.g., sonar), or compact single-axis scanning laser rangefinders to sense the floorplan of a room.
  • For terrestrial applications, robotic vehicles need to perceive both the 3-D geometry and the composition of the terrain (e.g., to discriminate traversable vegetation from nontraversable rocks). For some basic discriminations, viable solutions are in hand; in particular, live vegetation is easily distinguished from
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
  • soil and rocks using visible and near-infrared imagery (Matthies et al., 1996). Other discriminations are still poorly solved, especially for real-time applications, such as distinguishing dead vegetation from soil. For some of these cases we are studying the use of image texture for terrain classification.
  • In addition to obstacle detection, position estimation is a major part of the autonomous navigation problem. Although the global positioning system (GPS) can largely solve this problem outdoors on Earth, it is still an important problem for indoor robot applications and in planetary exploration. Visual feature tracking with stereo cameras has been employed successfully for robot motion estimation and for terminal guidance to human-designated objectives (Matthies and Shafer, 1987; Wettergreen et al., 1997). A number of methods are under development that use maps together with images and other sensors for various forms of landmark recognition (Cozman and Krotkov, 1996; Lu and Milios, 1997; Matthies et al., 1997). Some of these methods are fairly mature for Earth applications now; methods suitable for Mars rovers will likely come to maturity and be deployed over the next five to seven years.
  • In terrestrial applications, autonomous navigation among other moving objects requires a significant extension of perception, planning, and local world modeling capabilities beyond that addressed above. Much work is in progress on this problem, using sonar, scanning-laser rangefinders, and imagery.

Predictions

Robotic perception systems for Mars rovers should enable autonomous navigation up to a kilometer or more from the lander by the year 2003. Laser rangefinders and image-based feature tracking and landmark recognition algorithms are expected to be used for autonomous precision landing on a comet in less than 10 years from now. Within 10 years it is also possible that unmanned ground vehicles will be sufficiently mature to proceed with fielding them for selected military applications. The cost of such systems may also be low enough, and the capability high enough, to support commercialization for some civil applications. Potential commercial applications of such technology include autonomous material transportation and collision avoidance sensors for smart passenger vehicles. Applications for computer vision technology also exist outside autonomous navigation, such as in PC-based camera systems that use 3-D shape and motion-sensing capabilities to enhance video conferencing and human-computer interfaces. In fact, commercial imaging applications are part of what is driving the rapid progress in low-power CMOS imagers, low-power embedded processors, and vector processing extensions to general-purpose microprocessor architectures. These advances will lead to computer vision systems in the next 10 years that will make robotic vehicles and the vision systems themselves cost-effective for new applications. Finding and exploiting these markets will be an exciting opportunity for engineers in the next decade.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Acknowledgments

The research described in this paper was carried out by the Jet Propulsion Laboratory, California Institute of Technology, with support from the National Aeronautics and Space Administration, Defense Advanced Research Projects Agency, and the Joint Robotics Program of the Office of the Secretary of Defense.

References

Cozman, F., and E. Krotkov. 1996. Position estimation from outdoor visual landmarks for teleoperation of lunar rovers. Proceedings of the Third IEEE Workshop on Applications of Computer Vision . Los Alamitos, Calif.: IEEE Computer Society Press.


Dickmanns, E. D., and B. Mysliwetz. 1992. Recursive 3-D road and relative ego-state recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence 14(2):199-213.


Konolige, K. 1998. Small vision systems: Hardware and implementation. Proceedings of the Eighth International Symposium on Robotics Research, Y. Shohira and S. Hirose, eds. New York: Springer-Verlag.


Lu, F., and E. Milios. 1997. Robot pose estimation in unknown environments by matching 2D range scans. Journal of Intelligent and Robotic Systems 18:249-275.


Marr, D. 1982. Vision. San Francisco, Calif.: W. H. Freeman.

Matthies, L. H. 1992. Stereo vision for planetary rovers: Stochastic modeling to near real-time implementation. International Journal of Computer Vision 8(1):71-91.

Matthies, L. H., and S. A. Shafer. 1987. Error modeling in stereo navigation. IEEE Journal of Robotics and Automation RA-3(3):239-248.

Matthies, L., A. Kelly, T. Litwin, and G. Tharp. 1996. Obstacle detection for unmanned ground vehicles: A progress report . Proceedings of the Seventh International Symposium on Robotics Research, G. Giralt and G. Hirzinger, eds. New York: Springer-Verlag.

Matthies, L., C. Olson, G. Tharp, and S. Laubach. 1997. Visual localization methods for Mars rovers using lander, rover, and descent imagery. Proceedings of the International Symposium on Artificial Intelligence, Robotics, and Automation in Space (i-SAIRAS). Washington, D.C.: National Aeronautics and Space Administration.

Mettala, E. G. 1992. The OSD tactical unmanned ground vehicle program. Pp. 159-172 in Proceedings of the DARPA Image Understanding Workshop. San Mateo, Calif.: Morgan Kaufmann Publishers.


O'Handley, D. A. 1973. Scene analysis in support of a Mars rover. Computer Graphics and Image Processing 2:281-297.

Olin, K. E., and D. Y. Tseng. 1991. Autonomous cross-country navigation. IEEE Expert 6(4):16-32.


Wettergreen, D., H. Thomas, and M. Bualat. 1997. Initial results from vision-based control of the Ames Marsokhod rover. Proceedings of the International Conference on Intelligent Robots and Systems (IROS). New York: Institute of Electrical and Electronics Engineers.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Cobots

Michael A. Peshkin

Mechanical Engineering Department Northwestern University Evanston, Illinois

It is often assumed that the benefit of robots is their strength, speed, or accuracy—qualities that a human operator's "help" can only diminish and that indeed may pose a hazard to him or her. Less recognized is that it may be the robots' interface to computers and information systems that is their primary benefit. Furthermore, even in heavily mechanized environments, people have an important continuing role because of their sensing and dexterity, which cannot be matched or replaced by robots.

Collaborative robots—"cobots"—are a new type of robotic device, intended for direct interaction with a human operator in a shared workspace. Cobots allow a true sharing of control between human and computer. The human operator supplies motive power and exerts forces directly on the payload, while the mechanism of the cobot serves to redirect or "steer" the motion of the payload under computer control. The computer monitors the force (direction as well as magnitude) applied by the operator to the payload. In real time these operator forces can be compared with programmed guiding surfaces, and motion in the direction that the operator pushes can be allowed, disallowed, or redirected. The human operator may be allowed complete freedom of motion of the payload, or in the opposite extreme the payload may be restricted to a single curve through space. Thus, the full versatility of software is available for the production of virtual surfaces and other haptic effects.

At Northwestern, our first application area has been automotive assembly, with the help of General Motors. Unlike body welding and painting, automobile assembly is a worker-intensive process because of the need for human dexterity. Ergonomics issues have become a major concern: even payloads well within human strength limits for lifting can cause significant problems. In response, the

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

materials handling industry has developed a great variety of so-called "assist devices" that provide support against gravity and sometimes guide motion but do so at the expense of much greater inertia, anisotropic response to the operator's applied forces, restriction of motion to few dimensions, and greater possibilities of jamming in assembly operations. Thus, conventional assist devices usually reduce productivity. Their clunkiness frustrates operators, often leading to disuse in practice. Damaging collisions may occur, reducing productivity.

Virtual surfaces, whether enforced by a cobot or a conventionally actuated robot, can solve many of the problems of assist devices, if they can be implemented on an appropriately large scale of workspace size and strength. However, the force magnitudes needed are almost by definition at least those of humans. The speeds, if we are to increase rather than decrease productivity, must be at least those of humans. This would seem to imply a need for large motors with greater-than-human power. Workers are understandably leery of such motors in the context of a general-purpose manipulator, if they are intended to work within its workspace.

Cobots, in contrast, rely on the worker to provide motive power or can give some small powered assistance that requires only small motors. The much greater need for force is that required for changes of direction, sometimes called "inertia management." In cobots this is accomplished by the physical mechanism of the cobot rather than by motors, with a consequent improvement in both safety and smoothness of operation.

Space permits only the briefest explanation of the mechanism of cobots. The simplest cobot has a two-dimensional (planar) workspace and a mechanical heart that is a single wheel (see Figure 1). Most interest lies in extension of the cobot idea to many dimensions of motion and to revolute joints. The latter uses a continuously variable transmission in place of a wheel. Interested readers are referred to http://cobot.com for further information.

The motor in Figure 1 simply steers the wheel. No amount of malevolent steering by the control computer can cause the cobot to move on its own. Only the operator can cause it to move, by applying forces to the handle. A force sensor (top) monitors these user forces.

The unicycle cobot displays two essential behaviors: free mode and virtual surface. Free mode is invoked when the cobot's position in its planar workspace is away from all defined constraint surfaces. The cobot should therefore permit any motion the user attempts to impart. To do this, the steering angle of the wheel is servocontrolled such that user forces perpendicular to the wheel's rolling direction are nulled. The behavior is similar to that of a caster wheel on a rolling item of furniture, although there is no physical caster at all.

When the user brings the cobot's position in the plane to a place where a constraint surface is defined, control of the steering angle changes over to virtual surface mode. The wheel is steered such that its rolling direction becomes tangent to the constraint surface, and this tangency is maintained as the user moves

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

Figure 1 

The unicycle cobot is the simplest possible cobot.  Constraint surfaces, defined in software, delimit  excluded areas of the plane. The Cartesian rails  serve only to keep the unicycle upright and  are not needed in practical cobots of higher  dimensionality. The cobot mechanism consists  of a free-rolling wheel in contact with a working  surface. The wheel's rolling velocity is monitored by  an encoder but is not driven by a motor.

the cobot in "virtual contact" with the constraint surface. The user perceives contact with a hard frictionless constraint surface. In practice the illusion is convincing. Virtual surface mode is ended when the measured user forces are found to be directed away from the constraint surface, at which point free mode resumes.

The unicycle cobot can be generalized to higher dimensions as well as to the revolute architecture characteristic of most industrial robots. The latter rely on a distinct kinematic element in place of the wheel-a continuously variable transmission. Many other haptic effects are also possible: virtual paths and attractive surfaces.

In materials handling applications such as automobile assembly, even the simplest haptic effect-free mode-can be very useful. In free mode the cobot gives the operator the perception that the payload is responding in an unconstrained and natural way to his applied forces. This is actually a simulated lack of constraint, and the existence of a computer in the loop gives an opportunity for many improvements over the natural behavior of the payload-virtual haptic effects. For instance, the lack of isotropy of the underlying kinematic mechanism (e.g., an overhead rail system or an articulated arm) can be masked by the cobot in free mode, so that the payload responds in a more predictable way to the

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

operator's intentions. A prototype railcobot at Ford Motor Company includes this masking effect. Or the inertia of the payload—its reluctance to change its direction of motion—can be masked so that it is perceived as lighter and more maneuverable than it actually is.

Virtual surfaces are another haptic effect. To be useful they must be hard (abrupt), strong (large forces sustained without penetration), and smooth (no friction as a payload is pushed along a virtual surface). Since a cobot's virtual surfaces rely on a physical mechanism rather than on actuators, these desirable qualities are innate. Virtual surfaces are useful for productivity because an operator can push a payload against a virtual surface and "swoop" around a corner quickly and pleasantly.

Virtual surfaces can also be used to prevent collisions in close quarters or assembly operations. A cobot now under test at General Motors assists in removing doors from car bodies after painting and prior to assembly. It confines the door's motion (rotation as well as translation) to a well-chosen curved escape path over a few critical inches as it is removed from the car body, preventing collision of finished surfaces.

While materials handling and automotive assembly in particular have been the first applications area, many others exist as well.

  • Image-guided surgery. In this area safety is essential, and a cobot's ability to guide motion without possessing a corresponding ability to move on its own can totally remove concern about some failure modes. Perhaps more importantly, the quality of a virtual surface enforced by a cobot originates in its physical mechanism, rather than in servo-controlled actuators, thus yielding harder and smoother surfaces than can be achieved by a robot. Preserving the critical sense of touch in surgery requires high-quality "shared control" between surgeon and robot, for which smoothness of motion is essential.
  • Haptic display. Computer-aided design (CAD) models of contoured objects (e.g., beverage bottles, car bodies) can be displayed visually, but the feel of these objects cannot be experienced prior to building them. Cobots can display hard, smooth virtual surfaces from CAD models.
  • Rehabilitation and exercise. Popular weight training equipment, originally designed for rehabilitation, uses shaped cams and other mechanical components to confine a user's limb or body motion to a particular trajectory. While these trajectories are somewhat adjustable, far greater versatility could be achieved if the motion trajectories were encoded in software rather than frozen into the mechanical design of the equipment. Cobots can enforce virtual trajectories with the smoothness, hardness, and safety required for this application.

Many fascinating research areas have been exposed in building and controlling cobots. Many of the research topics that have been explored in robotics

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×

suggest new and different questions in the context of cobots. A sampling includes the following:

  • Path planning. Creating the appropriate motion trajectory for a given task is a current issue in robotics. In cobots the corresponding problem is to create the virtual surfaces that bound and guide the motion of a payload controlled by a human operator, in support of a given task.
  • Haptic effects. Free mode and virtual surface mode are but two poles of an unlimited range of haptic effects that can be invented. For instance, a virtual surface may have a "penetration strength" beyond which it gives way, or it may have a simulated attractive potential field, or it may yield compliantly to operator pressure against it.
  • High dimensions. For cobots with workspace dimension greater than two, virtual surfaces can exist with a variety of dimensionalities ("surface" remains the generic term). Describing these surfaces efficiently and usefully is nontrivial.
  • Control. Novel control issues are created by the essential role of the human operator in the motion of a cobot. For instance, a robot trajectory is a path through space parameterized by time. In cobot control, progress along a path may be entirely at the discretion of the human operator, who may stop or even reverse direction along a path. The utility of time as a parameter is thus greatly reduced, yet control software must maintain the cobot on the path.

Acknowledgments

The invention of cobots, by the author and colleague J. Edward Colgate, came about as part of a research project on intelligent assist devices funded by the General Motors Foundation. Further support by the National Science Foundation is gratefully acknowledged. For more information and bibliography references, please begin at http://cobot.com. Intellectual property rights to the cobot are held by Northwestern University.

Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
This page in the original is blank.
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 85
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 86
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 87
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 88
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 89
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 90
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 91
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 92
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 93
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 94
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 95
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 96
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 97
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 98
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 99
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 100
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 101
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 102
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 103
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 104
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 105
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 106
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 107
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 108
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 109
Suggested Citation:"Robotics." National Academy of Engineering. 1999. Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering. Washington, DC: The National Academies Press. doi: 10.17226/6411.
×
Page 110
Next: Dinner Speech »
Frontiers of Engineering: Reports on Leading Edge Engineering From the 1998 NAE Symposium on Frontiers of Engineering Get This Book
×
Buy Paperback | $51.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF
  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!