Biologically inspired robotics

• Dr. Randall D. Beer, Cognitive Science Program, Indiana University, Bloomington, IN, USA

Figure 1: A biologically-inspired underwater robot based on the structure of a manta ray. The robot is capable of both hydrostatic gliding and active propulsion by flapping its two large pectoral fins, with high maneuverability. Image courtesy of Dr. Rudolf Bannasch at EvoLogics.

Biologically-Inspired Robotics is part of a body of work at the intersection of biology and robotics. In order to avoid confusion, this article will use the term biorobotics to refer to this broad intersection, reserving the term biologically-inspired robotics for work whose primary focus is the application of biological ideas to address technological problems (sometimes also called biomimetics), and the term biorobotic modeling for work whose primary focus is the use of robots as a modeling tool for addressing biological questions.

History

The creation of artificial devices with life-like characteristics has been pursued for over 2000 years, beginning, as did so many things in our modern world, in Ancient Greece. For example, among the inventions of Hero of Alexandria were a windmill-operated pipe organ and a mechanical theatrical play. In the 13th century, the Arab scholar and craftsman Al-Jazari created a series of hydraulic automata, including a drink-serving waitress and a musical band consisting of four performing mechanical musicians floating in a boat. During the 18th and 19 century, mechanical puppets or dolls known as karakuri ningyō were popular in Japan. Also during the 18th century, the French engineer Jacques de Vaucanson constructed and exhibited a variety of automata, including a mechanical flute player and his famous mechanical duck. Another famous automata from the end of the 18th century was Maillardet's Draughtsman-Writer, which could produce three different handwritten poems and four different sketches. Wood (2002) provides a popular survey of the history of automata, and general information can also be found in the [Automaton] article on Wikipedia.

With the rise of cybernetic approaches in the late 1940s and early 1950s, a wide variety of electromechanical machines designed to mimic biological processes and systems were constructed (Cordeschi, 2002). Perhaps the best-known and most directly relevant to biorobotics is W. Gray Walter's robotic "tortoises" Elsie and Elmer (Walter, 1953/1963; Holland, 2003). Walter was a physiologist who made important early contributions to electroencephalography and clinical neurophysiology. His tortoises were small mobile robots covered by a hard shell. The robots were driven by steerable motorized wheels and possessed a headlight, a light sensor, and a touch sensor that responded when the shell was hit. Their behavior was controlled by electronic circuit analogues of neural circuits. The behavioral repertoire of the tortoises included exploration, both positive and negative phototropisms, and obstacle avoidance. The activation of these different behaviors in interaction with the robots' environment could produce a variety of behavioral sequences. Although originally designed to explore Walter's theories of brain function, the tortoises became objects of popular fascination in much the same way that ancient automata did.

The seeds of the modern renaissance of biorobotics were sown from the mid 1980s to the mid 1990s. A key event in this resurgence was Rodney Brooks' (review 1991) work on behavior-based robots. Although not as directly based on biology as later work would be, Brooks argued that nontrivial and flexible behavior in a robot could be generated by the interaction between simple control machinery and its environment, demonstrating his point with robots accomplishing such tasks as insect-like walking. Another important milestone was Raibert's (review 1986) work on hopping and legged robots, which emphasized the central role of energetics in the dynamic balance and locomotion of animals. Arkin (1990), building on earlier work by Michael Arbib, developed a control architecture for reactive robots based on schema theory. Based on studies of serpentine motion, Hirose (review 1993) developed a number of snake-like locomotors and manipulators. In the early 1990s, Beer, Quinn, Chiel & Ritzmann (review 1997) developed a series of hexapod robots based directly on the body morphology and neural control of cockroach and stick insect walking (the latter, like many hexpod robots, based on the work of biologist Holk Cruse). In addition, Triantafyllou & Triantafyllou (1995) built a series of very efficient swimming robots based on studies of the hydrodynamics of fish swimming. Early biorobotic work on the sensory side includes Franceshini's (1992) robotic compound eye based on studies of insect eyes and motion-sensitive neurons in the fly, Webb's (1995) robotic model of cricket phonotaxis, Grasso et al's (1996) robotic model of lobster chemical orientation strategies, and Sahabot (Lambrinos et al., 1997), whose navigation by polarized light compass was inspired by studies of homing behavior in the desert ant. An early example of robots whose control was based on theories of human brain function is given by the work of Edelman et al. (1992).

Key Principles

As mentioned above, biorobotics actually consists of two different endeavors, biologically-inspired robotics and biorobotic modeling. However, these two endeavors have distinct motivations and criteria for success. Thus, although it is certainly possible for a single research project to make both technological and scientific contributions, it is important to carefully distinguish which aspects of a biorobotic research project are intended as contributions to which field.

In biologically-inspired robotics, the primary goal is technological: Biologically-inspired roboticists wish to build better robots. They look to biology for inspiration because, compared to current robots, the behavior of animals is extremely flexible and robust in the face of environmental contingencies. The hope is that adopting some of the design principles of animals will endow robots with similar flexibility and robustness. Biological inspiration can be drawn from many aspects of animals, including their behavioral strategies, the physical design of their bodies, and the organization of their nervous systems. Many degrees of biological inspiration are also possible, from vague resemblance to strict emulation. Major issues that must be considered include the degree of realism necessary to reap the benefits of biological inspiration and the separation of incidental biological details from those essential to performance of the task of interest. Often, a roboticist will use ideas from biology as a springboard for new engineering designs, subsequently ignoring biological realism. This is as it should be since, as a technological endeavor, the success of a biologically-inspired robotics project must not be judged by its faithfulness to the biological data. Rather, it must be judged by the extent to which the performance of the biologically-inspired robot improves upon existing technological approaches using whatever performance metrics are standard for that technology.

In contrast, the primary goal of biorobotic modeling is scientific: Biorobotic modelers wish to understand the mechanisms of animal behavior. As such, like any other biological model, biorobotic models must be judged by the extent to which they account for and illuminate the observed behavioral and neurobiological data, as well as the extent to which they generate testable experimental hypotheses. It is becoming increasingly clear that the mechanisms of animal behavior must be sought not only in that animal's nervous system, but also in its body and environment and the dynamics of interaction between these three components (Chiel and Beer, 1997). The special advantage of biorobotic models over computational ones derives from the fact that accurately modeling the physical body and world of an animal can be extremely difficult and computationally expensive, whereas the physics comes "for free" in a physically-instantiated biorobotic model. However, a major difficulty in robotic modeling is ensuring that the relevant physical, sensory and motor properties of the robot sufficiently match those of the animal relative to the biological question of interest, for if they do not, then the robot might actually work against the model's biological relevance. Webb (2001) provides an in-depth discussion of these and other issues related to biorobotic modeling.

Two Examples

There has been an explosion of work in biorobotics in recent years, with robotic vocal tracts, jaws, retinas, expressive faces, hands, arms, legs, etc. deployed on robotic worms, snakes, ants, flies, crickets, cockroaches, walking stick insects, dinosaurs, bats, lobsters, tuna, pickeral, turkeys, apes and humanoids. In addition, there is closely related work on swarm robotics, developmental robotics, and evolutionary robotics. Thus, no brief introduction could possibly do justice to the range of work currently being undertaken. A variety of reviews can be found in (Venkataraman & Iberall, 1990; Hirose, 1993; Beer, Ritzmann & McKenna, 1993; Beer, Chiel, Quinn & Ritzmann, 1998; Webb, 2001; Webb & Consi, 2001; Ayers, Davis & Rudolph, 2002; Ritzmann, Gorb & Quinn, 2004; Seth, Sporns & Krichmar, 2005; Bekey, 2005). This section will be limited to brief descriptions of one example each of biologically-inspired robotics and biorobotic modeling.

Figure 2: RiSE climbing a tree. Image courtesy of Dr. Dan Koditschek

A recent example of biologically-inspired robotics is Spenko et al's (2008) work on a hexapedal robotic climber called RiSE ( Figure 2) which, like a number of biologically-inspired legged robots, is based in part on work by the biologist Robert Full. In order to grip a vertical surface, this robot combines both bonding mechanisms inspired by the structure of gecko feet and interlocking mechanisms inspired by the structure of insect spines and claws. In addition, its design is based on a set of principles that have been found to be common to many climbing animals: (1) a sprawled posture keeps the body close to the surface so as to reduce the pitch-back moment; (2) front limbs pull inward and rear limbs push outward so as to counteract the pitch-back moment; (3) a long body reduces the pull-in force required of the front limbs; (4) lateral forces act inward toward the central axis of the body; (5) compliant legs, ankles and toes so as to distribute contact forces. Each of the six legs of RiSE have two degrees of freedom and the robot also possesses a static tail that presses against the surface to reduce the pull-in forces required of the front legs. The robot uses a wave gait in which only one leg at a time is lifted from the surface. In addition to an open-loop gait generator, RiSE utilizes a variety of feedback controllers, including traction force control, normal force control and gait regulation. In addition, the robot has a pawing behavior that allows a foot that fails to grasp on initial contact to reestablish a grip on the climbing surface. Spenko et al. have demonstrated that RiSE is able to traverse a variety of horizontal and vertical surfaces, including climbing trees and brick or cinder block walls.

Figure 3: A biorobotic model of lift generation in the fruit fly Drosophila melanogaster. Image courtesy of Dr. Michael Dickinson.

A powerful example of biorobotic modeling is provided by the aerodynamics of insect flight. Although quasi-steady-state aerodynamical analyses of the sort used to understand aircraft have been successfully applied to larger animals, they have not been very successful for explaining the generation of lift in small flying insects due to the tiny wingspans, relatively slow flight speeds and extremely fast wing movements involved. However, a recent biorobotic model by Dickinson and colleagues has begun to shed considerable light on the unsteady aerodynamics insect flight (Dickinson et al., 1999; Figure 3). Because of the delicate size and high speed of insect wings, direct measurement of the forces involved is extremely difficult. For this reason, a robotic model with a 60 cm wingspan was used to explore the non-steady-state airflow during hovering by the fruit fly Drosophila melanogaster. In order to reproduce the Reynolds number (ratio of inertial to viscous forces) relevant to small insects flying in air, their model was submerged in mineral oil and scaled both in space and time. Force sensors at the base of one wing allowed direct measurement of the forces produced and illumination of air bubbles in the tank allowed direct observation of the fluid flow around the robotic wings. Dickinson and colleagues found that three major mechanisms contributed to lift generation in the model. First, vortices formed at the leading edge of the wing produce lift during much of the power stroke. Second, additional lift is produced by circulation of air around the wings due to rapid rotation at the beginning and end of each stroke. Third, further forces are produced at the start of each upstroke and downstroke due to collisions of the wings with the swirling wake produced by the previous stroke, a mechanism termed wake capture. Due to the sensitivity of the latter two mechanisms to the timing of wing rotation, the model suggests that the control of small details of wing motion can be used in steering flight.

References

• Arkin, R.C. (1990) Integrating behavioral, perceptual and world knowledge in reactive navigation. Robotics and Autonomous Systems, 6:105-122

• Ayers, J., Davis, J.L. and Rudolph, A., Eds. (2002) Neurotechnology for Biomimetic Robots. MIT Press

• Beer, R.D., Quinn, R.D., Chiel, H.J. and Ritzmann, R.E. (1997) Biologically-inspired approaches to robotics. Communications of the ACM, 40:30-38

• Beer, R.D., Chiel, H.J., Quinn, R.D. and Ritzmann, R.E., Eds. (1998) Biorobotic approaches to the study of motor systems. Current Opinion in Neurobiology, 8:777-782

• Beer R.D., Ritzmann R.E. and McKenna T. (1993) Biological Neural Networks in Invertebrate Neuroethology and Robotics. Academic Press

• Bekey, G.A. (2005) Autonomous Robots. MIT Press

• Brooks, R.A. (1991) New approaches to robotics. Science, 253:1227-1232

• Chiel, H.J. and Beer, R.D. (1997) The brain has a body: Adaptive behavior emerges from the interactions of nervous system, body and environment. Trends in Neurosciences, 20:553-557

• Cordeschi, R. (2002) The Discovery of the Artificial: Behavior, Mind and Machines Before and Beyond Cybernetics. Springer.

• Dickinson, M.H., Lehmann, F.O. and Sane, S.P. (1999) Wing rotation and the aerodynamic basis of insect flight. Science, 284:1954-1960

• Edelman, G.M., Reeke, G.N., Gall, W.E., Tononi, G., Williams, D. and Sporns, O. (1992) Synthetic neural modeling applied to a real-world artifact. Proc. Natl. Acad. Sci. 89:7267-7271

• Franceschini, N., Pichon, J.M. and Blanes, C. (1992) From insect vision to robot vision. Phil. Trans. R. Soc. Lond. B., 337:283-294

• Grasso, F., Consi, T., Mountain, D. and Atema, J. (1996) Locating odor sources in turbulence with a lobster inspired robot. In P. Maes, M. Mataric, J.-A. Meyer, J. Pollack and S. Wilson (Eds.), From Animals to Animats 4: Proc. Fourth Intl. Conf. on Simulation of Adaptive Behvior (pp. 104-112). MIT Press

• Hirose, S. (1993) Biologically Inspired Robots: Snake-Like Locomotors and Manipulators. Translated from Japanese by P. Cave and C. Goulden. Oxford University Press

• Holland, O. (2003) Exploration and high adventure: The legacy of Grey Walter. Phil. Trans. R. Soc. Lond. A, 361:2085-2121

• Lambrinos, D., Maris, M., Kobayashi, H., Labhart, T., Pfeifer, R. and Wehner, R. (1997) An autonomous agent navigating with a polarized light compass. Adaptive Behavior, 6:131-161

• Raibert, M.A. (1986) Legged Robots that Balance. MIT Press

• Ritzmann, R.E., Gorb, S.N. and Quinn, R.D., Eds. (2004) Arthropod locomotion systems: From biological materials and systems to robotics. Special Issue of Arthropod Structure & Development, 33(3)

• Seth, A.K., Sporns, O. and Krichmar, J.L. (2005) Neurobotic models in neuroscience and neuroinformatics. Neuroinformatics, 3:167-170

• Spenko, M.J., Haynes, G.C., Saunders, J.A., Cutkosky, M.R., Rizzi, A.A., Full, R.J. and Koditschek, D.E. (2008) Biologically inspired climbing with a hexapedal robot. Journal of Field Robotics, 25:223-242

• Triantafyllou, M.S. and Triantafyllou, G.S. (1995) An efficient swimming machine. Scientific American, 272(3):64-70

• Venkataraman, S. and Iberall, T., Eds. (1990) Dextrous Robot Hands. Springer.

• Walter W.G. (1963) The Living Brain. W.W. Norton. This is a reprint of a book that was originally published in 1953

• Webb, B. (1995) Using robots to model animals: A cricket test. Robots and Autonomous Systems 16:117-134.

• Webb, B. (2001) Can robots make good models of biological behaviour? Behavioral and Brain Sciences, 24:1033-1050

• Webb, B. and Consi, T.R., Eds. (2001) Biorobotics: Methods and Applications. AAAI Press

• Wood, G. (2002) Edison's Eve: A Magical History of the Quest for Mechanical Life. Anchor Books

Internal references

• Valentino Braitenberg (2007) Brain. Scholarpedia, 2(11):2918.

• Randall D. Beer and Hillel J. Chiel (2008) Computational neuroethology. Scholarpedia, 3(3):5307.

• Max Lungarella (2007) Developmental robotics. Scholarpedia, 2(8):3104.

• Paul L. Nunez and Ramesh Srinivasan (2007) Electroencephalogram. Scholarpedia, 2(2):1348.

• Rodolfo Llinas (2008) Neuron. Scholarpedia, 3(8):1490.

• Jeff Krichmar (2008) Neurorobotics. Scholarpedia, 3(3):1365.

• John Dowling (2007) Retina. Scholarpedia, 2(12):3487.

See also

Neurorobotics, Swarm Robotics, Developmental Robotics, Evolutionary Robotics, Neuroethology, Computational Neuroethology, Animats

Sponsored by: Eugene M. Izhikevich, Editor-in-Chief of Scholarpedia, the peer-reviewed open-access encyclopedia

Sponsored by: Prof. Jan Peters, Max-Planck Institute, Germany & University of Southern California, USC

Reviewed by: Anonymous

Reviewed by: Dr. George Bekey, U. of Southern California

Accepted on: 2009-03-19 16:49:29 GMT