Ayers, J. (1995) A Reactive Ambulatory Robot Architecture for Operation in Current and Surge. In: Proc. of the Autonomous Vehicles in Mine Countermeasures Symposium. Naval Postgraduate School. Pp. 15-31

A Reactive Ambulatory Robot Architecture for Operation in Current and Surge

Joseph Ayers

Department of Biology and
Marine Science Center
Northeastern University
East Point, Nahant, MA 01908
(781) 581-7370
FAX: (781) 581-6076
Internet: lobster@neu.edu


Animals have evolved to occupy every environmental niche where we might want an autonomous robot to operate, save outer space. As a result, they provide proven solutions to the problems of navigation, searching and sensing in the most difficult of environments. Much proposed robotic behavior has precise analogies in the behavior of animals. In crustaceans, many action patterns can be evoked by stimulation of single neurons or sets of neurons, peptides and/or amines. Thus there is a correspondence between units of behavior, their modulation and underlying neuronal components. The organizational principles of these components can form the basis for a conservative object-based finite-state machine architecture which may apply to a variety of robotic systems. In this manuscript, I apply examples of this approach to the design of a behavioral controller for a reactive lobster-based underwater robot.

Fig. 1 Lobster compensatory postures to low (upper panel) and high speed (lower panel) water currents.


One strategy for mine clearance in the littoral region is to use large numbers of relatively inexpensive robots which crawl around on the bottom in a semi-random fashion. The near-shore region of the ocean presents a challenging environment for such a work effort. Heavy turbulence caused by surge, wave action, tides and currents causes severe stability problems and limits the range of required sensors. Moreover, the requirement of locating both projecting as well as buried mines requires highly adaptable sensors. These problems have been overcome by bottom-dwelling organisms such as crabs and lobsters who live with impunity in these complex conditions (Atwood and Sandeman, 1983; Cobb and Phillips, 1980). Their ambulatory locomotory movements, combined with hydrodynamic adaptability are a proven solution to the stability problem posed by this environment. In this manuscript we will argue that an ambulatory robot, based on the lobster, can be realized with available technology and share these traits.

A lobster-based robot is superior for operation in shallow complex inshore environments which feature currents and surge. A hydrodynamically adaptable shape can hold a lobster-based robot to the bottom rather than allowing current surge to displace it (Fig. 1). Lobsters and crayfish use their claws, abdomen and swimmerets as hydrodynamic control surfaces and thrusters affording considerable adaptation relative to hydrodynamic perturbation during locomotion (Davis, 1968; Maude and Williams, 1983; Ayers and Schlichting, 1995). Furthermore, their ability to walk adaptively in any direction allows them to preserve these hydrodynamic advantages like a wind vane while participating in a search procedure on an arbitrary heading. An ambulatory robot solves the stability problems of floating vehicles. It is continuously in mechanical contact with the sea floor, so it can pan and scroll through postural changes and thereby stabilize sensors. Lobsters can readily navigate over irregular bottom types and around obstacles such as rocks, crevices and seaweed. Lobsters flourish in the benthic and littoral environments and have developed robust control systems for locomotion, sensing, searching, and consummatory behavior. These control systems present a proven solution and an optimized strategy for an underwater walking machine.

Rhythmic systems in crustaceans are among the best understood in the animal kingdom. The central neuronal mechanisms underlying locomotion were first established by study of simple animals including decapod crustacea, insects and annelids (Hoyle, 1976; Kennedy and Davis, 1977). These mechanisms have been formalized into a general model, the command neuron, coordinating neuron, central pattern generator model (CCCPG model) which has been demonstrated to be conserved throughout the animal kingdom (Herman et al., 1976; Evoy and Ayers, 1981; Kennedy and Davis, 1977; Stein, 1978). The model is composed of five major classes of components including central pattern generators (Pinsker and Ayers, 1983; Selverston and Moulins, 1987), command systems (Kupferman and Weiss, 1978), coordinating systems (Stein, 1976), proprioceptive and exteroceptive sensors (Wiersma and Roach, 1977) and phase and amplitude modulating sensory feedback (Stein, 1978).

We submit that biologically-based reverse engineering of CCCPG-based control systems is the most effective procedure to design autonomous underwater robot controllers as well as to establish detailed higher order control schemes for remote sensing procedures. The control schemes by which a lobster searches for and acquires prey provide excellent solutions to the problem of how an underwater robot can successfully search for and investigate mine-like objects.

Biology of Lobster Walking

Limb Movements
Lobster walking movements are performed around three major limb joints (Fig. 2): the thoraco-coxal joint generates protraction and retraction movements, the coxo-basal joint generates elevation and depression movements while the mero-carpopodite joint generates extension and flexion movements (Ayers and Davis, 1977a; Ayers and Clarac, 1978). Cyclic elevation and depression movements of the coxo-basal joints underlay the swing and stance phase movements respectively for walking in all four directions (Fig. 2). Propulsive forces are generated by movements of the thoraco-coxal and mero-carpopodite joints. Propulsive forces which are synergistic with depression for walking in one direction, become antagonistic for walking in the opposite direction (Fig. 2). The lobster step cycle consists of three phases: an early swing phase lifts the limb tip toward the initial position of the stance, an early stance phase drops the leg to initiate the stance and during the late stance phase the limb applies propulsive force and compensates for gravity. The swing phase is always constant in duration and variation in the step period is mediated by variation in the duration of the stance phase (Ayers and Davis, 1977a).

Fig. 2 Joint movement and muscle synergy control patterns. The three panels at left indicate the angular coordinates of movement of the coxo-basal joint (elevation and depression) , the thoraco-coxal joint (protraction and retraction) and the mero-carpopodite joint, (extension and flexion). The three upper graphics on the right indicate the angles of the coxo-basal and thoraco-coxal joints and mero-carpopodite plotted vs phase in the step cycle. Forward walking is indicated by closed circles while backward walking is indicated by open circles. The lower right panel indicates the duration of the swing phase and stance phase (stippled bars) and the corresponding elevation, depression and propulsion phases.

Neuronal Circuits Controlling Locomotion in Crustacea
The bulk of the macruran behavioral repertoire can be evoked by stimulation of single descending command neurons (Bowerman and Larimer, 1974a, b; Hoyle, 1976). A hypothetical neuronal network for the control of walking based on neuronal oscillators, command and coordinating neurons has been developed (Fig. 3a). This simple network is capable of controlling walking in four directions (Ayers and Davis, 1977a; Ayers and Crisman, 1992a; Ayers and Schlichting, 1995). The basis of the network is direct connections between oscillator neurons and propulsive force neurons (Chirachri and Clarac, 1989) modulated by command systems (Cattaert, et. al., 1990). According to this model, command systems specify the coordination pattern through modulation of specific connections to specify walking direction (Fig. 3a). Command systems also specify the amplitude of movements through recruitment from the motor unit pool (Davis and Kennedy, 1972; Ayers and Crisman, 1992a).

Fig. 3. Hypothetical neuronal network for the control of omnidirectional stepping. Closed circles indicate presynaptic or conventional inhibitory connections. Open triangles indicate excitatory connections. Fig 3B. Message hierarchy of the finite state machine. State change messages pass down the hierarchy. See text for explanation.

Lobster-Based Ambulation Controller

At present, we have completed the low level components of the omnidirectional ambulation controller and are currently implementing controls for behavioral sequences (Ayers and Schlichting, 1995). We implemented the CCCPG model for omnidirectional walking as a finite state machine. The ambulation controller program generates digital output for actuator control and real-time chart displays of motor programs imitating electromyograms. The program treats the CCCPG model components as objects which pass messages with regard to status changes. Changes in walking direction, speed, load etc., are effected by menu selections or key strokes. The program generates chart displays which are the basis of many of the figures of experimental results in this manuscript.

At the single limb control level, the ambulation controller relies on three major classes of components which control the elevator, depressor, protractor, retractor, extensor and flexor synergies (Fig. 3b). The oscillator component is a software clock which regulates the period of stepping as well as the duration of the swing or elevator phase fraction of the stepping cycle.

The second major component of the ambulation controller is the coordinator which determines the pattern of discharge of bifunctional synergies. The coordinator responds to the state transition message and desired period parameter from the oscillator; polls the walking command logic and determines, through a truth table, which synergies should be active; and sets or clears the booleans associated with different bifunctional synergies. The truth table implements the presynaptic inhibitory logic of the neuronal circuit model and specifies the excitatory connections which would be disabled by the directional command.

The third major component of the ambulation controller is the recruiter which determines which of the elements of the propulsive force synergies are active (Fig. 3b). The recruiter responds to the desired period parameter as well as load sensitive feedback and sets or clears the booleans associated with different elements of the active propulsive pool. In the current implementation, each unit within a synergy is represented as a boolean although it could easily be represented as a scalar to provide more detailed amplitude control.

Fig. 4 Motor programs for omnidirectional walking generated by the finite state machine. . Note the co-activation of antagonists for joints which serve a postural function. Ele: elevator synergy, Dep: depressor synergy, Pro: protractor synergy, Ret: retractor synergy; Ext: extensor synergy, Flx: flexor synergy. The time marks occur once/second.

In addition to controlling forward and backward walking, the ambulation controller implements omnidirectional locomotion including lateral walking on the leading and trailing sides (Fig. 3a). To complete this orthogonal symmetry, we have included directional and recruiting logic for backward, lateral leading and lateral trailing walking. In all cases, the directional logic eliminates the inappropriate synergies by decoupling propulsive force synergies from the elevation/depression oscillation and adds speed dependent recruitment to the propulsive force synergy. The added connections include those which decouple elevation from flexion and depression from extension during lateral leading walking and those which decouple elevation from extension and depression from flexion during lateral trailing walking.
The output of the finite state machine is control signals which specify the timing and amplitude of actuator action. Fig. 4 indicates the signals controlling the thoraco-coxal, coxo-basal and mero-carpopodite joints during walking in all four directions. Note that movement synergies which are synergistic for walking in one direction become antagonistic for walking in the opposite direction.. Adaptation to speed is mediated both by increases in the frequency of the pattern as well as recruitment of the propulsive synergy (Ayers and Crisman, 1992);

Crustacean sensors code environmental parameters in terms of labeled lines (Bullock, 1978). Each sensor is represented by an array of labeled line elements, each of which codes for a particular sensory modality (gravity, water current, etc) as well as a receptive fields (i.e. orientation relative to horizontal, water currents from the front, rear or the sides, etc.). In the controller implementation, sensor objects pass messages to command objects based on the actual input. Thus the message contains the modality and orientation and evokes different methods from the command system, based on these parameters. During locomotion in currents and surge, lobsters predominately rely on three sensors, the antennae, water current receptors and the statocysts or vestibular receptors. Although the actual biological sensors are complex, they can be readily modeled by mechanical transducers to code environmental information in the same fashion as the lobster nervous system.

In a robotic implementation antennae require both active motor control as well as a set of labeled lines which code for contact as well as current forces. In the simplest implementation, the antennae should be constructed of a tapered tube of moderate restoring force. Antagonist actuators would mediate protraction and retraction in the yaw plane. Strain gauges placed at the distal end would coded for contact while others placed at the basalar portion could code for water currents as well. Crustaceans perceive water currents using a combination of antennae as well as specialized directional sensors called hair fan organs which are distributed over the carapace and thorax. (Laverack, 1962). The hair fan organs consist of a cluster of hairs which can pivot in one plane (Fig. 5a). Water currents deflect hair pegs where the pivot of the hairs is oriented at right angles to the direction of current, so the organs provide labeled lines which can code both amplitude and direction. In a robotic implementation, hair fans could be implemented by strain gauges distributed over the "claws" (Fig. 5b).

Crustaceans perceive their orientation relative to gravity with a specialized organ the statocyst (Fig. 5c), which is located in a pit at the base of the antennules (Cohen, 1955). The statocyst contains receptors, statolith hairs which discharge when the statocyst is at different orientations relative to gravity. Statolith hairs exhibit range fractionation in the roll and pitch planes, with each hair coding for a particular orientation. A biologically-based implementation of a statocyst might consist of a rotating pendulum with slots at different orientations upon which state transition messages would be passed when switched between photodiode-based labeled lines (Fig. 5d).

A. B.

Fig. 5 A. Water Current Sensors. A. Hair Fan organ of macruran decapods (After Laverack, 1962). The central hair cluster is deflected b currents to provide labeled line for current in different directions. B. Analogous sensors on a robotic "claw". Each sensor element would consist of a small tube with a strain gauge inserted into the lumen. The housing would constrain the element to currents in a particular orientation and protect the gauge from mechanical insult. The individual elements would be placed on the "cheliped" hydrodynamic control surfaces at different angles. Simple thresholding of the strain gauge response would indicate different levels of current amplitude while the identity of the sensor would indicate the direction of current flow. The sensor array would then provide an array of labeled lines coding both direction and intensity. C. A. The lobster statocyst is located in a pit at the base of the antennules (After Cohen, 1955) . D. A biologically-based inclinometer. A pendulum would rotate between photodiode and detector arrays. Each detector would code for inclination over a particular range of orientations relative to gravity. Each photodiode would constitute a a labeled line would code for inclination at a particular angle. Two inclinometers in the roll and pitch planes would code in an analogous fashion to the crustacean statocyst.

Units of Behavior
Action Components: The atom of structure in the behavioral sequencer is the action component. Action components form the basic unit of transition and consist of the event type (forward, stand, etc), the previous event at the time the action pattern is evoked, the duration of the event, the latency of the event (for developing sequences) and the intensity of the event. The intensity corresponds to underlying parametric modulation in terms of amplitude and frequency. We employ two other modulatory components, 1) the speed for locomotory behavior, and 2) the posture in the roll, pitch and yaw planes. The intensity is analogous to parametric neuromodulation. It has a rise time, a fall time and a peak amplitude. These action components constitute the instruction set of the robot.

Action components are sequenced through three structures which correspond to the three major behavioral schemes: reflexes, modal action patterns and goal achieving behavior. In reflex patterns the duration and intensity are proportional to the intensity of the stimulus. Reflexes can provide continuous modulation of evoked or ongoing behavior. Exteroceptive reflexes respond to external stimuli and modulate command systems. Proprioceptive reflexes operate at the segmental level on individual limbs.

Compensatory Reflexes
Lobsters adapt to the environment through both segmentally mediated proprioceptive reflexes (Ayers and Davis, 1977b; Ayers and Crisman, 1992) as well as exteroceptive reflexes.Exteroceptive reflexes act on command systems and or intersegmental modulatory interneurons. An example of an exteroceptive reflex during locomotion is rheotaxis. In this behavior, the releaser is water currents directed from the sides perceived by hair fan organs or their biologically based analogs (Fig. 5b). Where small deviations from the midline are perceived the response is a yaw correction while larger deviations (>45š) would cause an intercalated rotation to the left or right (Fig. 6). These yaw correcting responses are followed by pitch compensation to provide a thrust vector toward the substrate (Fig. 6).

Fig. 6 Rheotaxic exteroceptive reflexes resulting from water currents from the left or right. The time of the stimulus is indicated by a vertical arrow. In both cases, the response is to rotate into the direction of the current as indicated by the diagrams at the right. Lower Panel: Postural compensatory responeses of a lobster walking from right to left in a laminar flume. Note how the chelipeds are lowered and the abdomen elevated.

Compensatory reflex responses to water currents involve simultaneous compensation in the pitch plane as well as postural responses of the abdomen and chelipeds. The ambulation controller supports an antigravity recruiter which acts on the depressor synergy of each leg and mediates pitch and roll compensation. During medium currents the controller reduces the depression in anterior segments which will pitch the hull forward. During high currents the controller also increases depression in the caudal segments causing even greater forward pitch. These compensatory responses to water currents and surge involve both yaw and pitch plane components as well as load compensation for the necessary added propulsive thrust.

The yaw components of navigation are mediated reactively by taxes and kineses (Loeb, 1918; Braitenberg, 1984). Positive yaw taxes or attraction occur when sensor bias directs locomotion toward a source or up a concentration gradient and are generally mediated by contralateral causality between sensors and effectors (Fig. 7). Negative yaw taxes or avoidance occur when sensor bias directs locomotion away from source and are generally mediated by ipsilateral causality between sensors and effectors (Fig. 7). In the ambulation controller, such inputs send messages to both the parametric command as well as the recruiting component. In attractive reflexes the messages is sent to the contralateral command objects while during avoidance reflexes (negative taxis) the messages are sent to the ipsilateral command objects. As a result, the controller both biases the period on the two sides (walking faster on the side turned away from) as well as the recruiters on the two sides to generate more propulsive force on the faster side. Control schemes like that of Fig. 7 can be easily adapted to a variety of beacon tracking, avoidance, docking and search behavior through acoustic, magnetic, optical, tactile, gravitational, flow and chemical sensors.

Figure 7. Networks for yaw plane exteroceptive reflexes. A. Positive Yaw Taxis. . B. Negative yaw taxis.. C. Yaw plane modulating motor patterns during forward walking. The traces indicate the elevator, depressor and retractor synergies of the left and right sides.

Crustacean muscle man be modeled mechanically as a linear actuator (sliding filaments) with a parallel elastic component (cell membranes) as well as a series elastic component (ligaments and tendons). Muscle contraction is typically activated by trains of action potentials which can cause graded or twitch contractions depending on the size of the motor units. (Atwood, 1976, Davis, 1971). An excellent candidate as actuators for a lobster-based robot are shape memory alloys. Shape memory metals undergo a state transformation from a deformable state (martensite) to a remembered state (austentite). When formed into wires, one such shape memory metal (nitinol) may change its length by as much as 10% during contraction (Gilbertson, 1992). Nitinol has a maximum and minimum range of action specified by its stretched and remembered lengths. Within this range its control is proportional, depending upon the level of current passed or the linear extent of the wire through which current is passed.

Fig. 8. Actuator models for an Ambulatory Robot. Left Panel: Artificial muscle based on nitinol ware arrays. The plastic sliders would support an array of 4 gauges of wire around a metal tube. One slider would be connected with a series spring. The different wire gauges correspond to different recruitment levels in the control signal.

Nitinol wires can be activated by the pulse trains generated by the ambulation controller much as crustacean muscle is activated by neuromuscular transduction (Atwood, 1976) and serve as the linear actuator for artificial muscle (Fig. 8). Different wire gauges can realize force recruitment in an analogous fashion to the recruitment strategies of muscle (Davis, 1971). These properties map directly on the architecture of the ambulation controller.

Building a Lobster-Based Robot.
Based on the architecture described above, it is quite feasible to implement a lobster-based robot using known components. We advocate adopting the basic body form of the lobster (cf. Figs 2, 9). The hull would be water-tight and contain the control circuitry, batteries and current drivers. Bilaterally paired "claws" would function as hydrodynamic control surfaces but could also support sensors and/or small charges for mine neutralization. The ability of a lobster-based robot to place charges in close proximity to a mine candidate would profoundly reduce the size of a necessary charge and their mass could be compensated for with a buoyant component. A tail structure with controllable "uropods" would perform an analogous function to the abdomen of lobsters. In addition it could easily mediate righting by elevating and bending to either side to produce righting torque (Davis, 1968).

Fig. 9. Proposed realization of a lobster-based robot.

The control circuitry of the robot could easily be realized on a small single board computer. The controller as presently implemented features few floating point calculations and both the sensor and effector signals are represented as arrays of bits. For example, an inclinometer is represented by one byte, each bit of which represents angular inclination over a particular range of angles. Similarly the shape memory actuators consist of an array of wires of different gauges, each element of which corresponds to a bit. Thus the control of each joint would require a byte. These input and output bytes can readily be implemented by two series of shift registers which are interfaced to the computer through serial line drivers. As the controller program executes it would both read from and write to these shift registers as each state transition occurs. During quiescence, the program might poll the sensor serial line to look for changes in the input or react to sonar control signal generated by a human operator. We have developed a transponder system for acoustic telemetry which could be readily adapted to messaging two and from shallow water robots (Massa, Ayers and Crisman, 1992).

The power requirement of such a robot will be relatively high and will depend largely on environmental factors. Due to relatively neutral buoyancy, a lobster-based robot needs to expend energy primarily to overcome the hydrodynamic resistance. Where currents are low and/or the substrate regular, the system can rely on the smallest actuator elements which draw the least current. As conditions worsen and stronger elements are needed to overcome hydrodynamic flow the power requirements will increase. As argued elsewhere (Jalbert, Kashin and Ayers, 1995) commercially available lithium technology batteries can provide adequate current for missions of several hours duration.

Fig. 10 Sensor/Actuator interfacing to single board computer. Labeled line code of biologically based sensors will consist of bits for each element of the range which are mapped on shift registers. The actuator signals will also consist of bits which are also mapped on shift registers. Bits in the output shift registers will control current driver circuits for each of the elements of a nitinol wire array.

It is feasible to base the design of an autonomous underwater robot on biological principles. Sensors, controlling circuits and actuators can readily be designed which operate on the same principles as their living analogs. Nature is conservative in the neuronal control strategies throughout the animal kingdom relying on the command system, coordinating system, central pattern generator model (Stein, 1978). It is worthwhile to explore in the future whether such a control architecture might also prove generalizable to different types of underwater robots (Ayers et al., 1994)

  • Atwood, H.L. (1976) Organization And Synaptic Physiology Of Crustacean Neuromuscular Systems. Prog. in Neurobiol. 7: 291-391
  • Ayers, J. (1992) Desktop Motion Video for Scientific Image Analysis. Advanced Imaging 7: 52-55.
  • Ayers, J. and F. Clarac. (1978) Neuromuscular strategies underlying different behavioral acts in a multifunctional crustacean leg joint. J. Comp. Physiol. 128: 81-94.
  • Ayers, J. and Crisman. J. (1992) The Lobster as a Model for an Omnidirectional Robotic Ambulation Control Architecture. In: Biological Neural Networks in Invertebrate Neuroethology and Robots., R. Beer., R. Ritzmann and T. McKenna [eds], 287-316.
  • Ayers, J. and W. J. Davis. (1977) Neuronal control of locomotion in the Lobster, Homarus americanus. I. Motor programs for forward and backward walking. J. Comp. Physiol. 115: 1-27.
  • Ayers, J. and Davis, W. J. (1977) Nervous control of Locomotion in the Lobster, Homarus americanus. II. Types of walking leg reflexes. J. Comp. Physiol 115: 29-46.
  • Ayers, J and Fletcher, G. (1990) Color Segmentation and Motion Analysis of Biological Image Data on the Macintosh II. Advanced Imaging 5: 39-42
  • Ayers, J. and Schlichting, L. (1995) Units of Behavior for a Lobster-Based Underwater Robot. Adaptive Behavior, in review
  • Ayers, J. and Selverston, A. I. (1979) Monosynaptic entrainment of an endogenous pacemaker network: a cellular mechanism for von Holst's magnet effect. J. Comp. Physiol. 129: 5-17.
  • Beer, R.D. (1991) Intelligence as adaptive behavior: An experiment in computational neuroethology. Academic Press. New York
  • Beer, R. D. Chiel, H.J. Sterling, L.S. (1993) Computer-simulated insects that adapt to their environment may be the next stage in the evolution of artifical intelligence. In: An Artificial Insect . American Scientist. Cleveland, OH. Pp. .
  • Bowerman, R. F. and Larimer,J.L; (1974a) Command fibres in the Circumoesophageal connectives of crayfish I. Tonic fibres. J. Exp. Bio. 60: 95-117.
  • Bowerman, R. F. and Larimer, J.L. (1974b) Command fibres in the circumoesophageal connectives of crayfish II. Phasic fibres. J. Exp. Biol. 60: 119-134.
  • Braitenberg, V. (1984) Vehicles: Experiments in Synthetic Psychology, MIT Press, Cambridge.
  • Brooks, R.A. (1991) New approaches to robotics. Science 253: 1227-1232.
  • Bullock, T. H. (1978) An Introduction to Neuroscience, Freeman, San Francisco.
  • Cattaert, D. ElManira, E. Marchand, A. and Clarac, F. (1990) Central control of the sensory afferent terminal from a leg chordotonal organ in crayfish in vitro preparation. Neuroscience Letters 108: 81-87.
  • Chrachri, A. and F Clarac. (1989) Synaptic connections between motor neurons and interneurons in the fourth thoracic ganglion of the crayfish, Procambarus clarkii. J. Neurophysiol. 62(6): 1237-1250.
  • Clarac, F. and W.J.P. Barnes. (1985) Peripheral influences on the coordination of the legs during walking in decapod crustaceans. Pp. 249-269. In: Coordination of Motor Behavior. B.M.H. Bush and F. Clarac [eds]. Cambridge University Press, Cambridge.Pp. 249-269.
  • Cobb, J.S.; Phillips, B.F. (1980) The Biology and Management of Lobsters, Volume II. Academic Press.
  • Cohen, M. J. (1955) The Function of Receptors in the Statocyst of the Lobster,(Homarus americanus). J. Physiol. 130: 9-33.
  • Davis, W. J. (1968) Lobster righting responses and their neural control. Proc. Roy. Soc. B. 70: 435-456.
  • Davis, W. J. (1971) Functional Significance of Motor Neuron Size and Soma Position in the Swimmeret System of the Lobster. J. Neurophysiology 34: 274-288.
  • Davis, W. J. and J. Ayers (1972) Locomotion: Control by positive feedback optokinetic responses. Science 177: 183-185
  • Davis, W. J. and Kennedy, D. (1972) Command Interneurons Controlling Swimmeret Movements in the Lobster 1. Types of Effects on Motoneurons. J. Neurophysiology 35: 1-12.
  • Delcomyn, F. (1980) Neural basis of rhythmic behavior in animals. Science 210: 492-498.
  • Eaton, R.C.; DiDomenico, R. (1985) Command and the Neural Causation of Behavior: A Theoretical Analysis of the Necessity and Sufficiency Paradigm. Brain Behav. Evol. 132-1624: 132-164.
  • Evoy, W. H. and Ayers, J. (1981) Locomotion and Control of Limb Movements. Chapter 16, In: The Biology of Crustacea (D. E. Bliss, ed.). Pp. 61-105.
  • Forssberg, H. (1979) Stumbling corrective reaction: A phase dependent compensatory reaction during locomotion. J. Neurophysiol. 42: 936-953.
  • Gilbertson, R. (1992) Motorless motion: working with shape memory wires Second Edition. Mondotronics, San Anselmo.
  • Herman, R.M. , S. Grillner, P.S.G. Stein, and D.G. Stuart. (1976) Neural Control of Locomotion. Plenum Press. New York.
  • Hoyle, G.[ed]. (1976) Identified Neurons and Behavior of Arthropods. Plenum Press.
  • Jalbert, J. Kashin, S. and Ayers, J (1995) Biologically-based Undulatory Lamprey AUV. This Volume.
  • Kennedy, D. and Davis, W.J. (1977) Organization of invertebrate motor systems.In: Handbook of Physiology. Geiger, R., Kandel, E.
  • Brookhart, J.M.[eds]. American Physiology Society. Bethesda Md,. Pp. 1023-1087.
  • Kupferman, I. and K. R. Weiss (1978) The command neuron concept. Behav. Brain Sci. 1: 3-39.
  • Laverack, M.S. (1962) Responses of cuticular sense organs of the lobster, (Homarus Vualgaris) (Crustacea)-1. Hair-peg organs as water current receptors. Comp. Biochem. Physiol. 5: 319-325.
  • Libersat, F., Clarac, F. and Zill, S. (1987) Force sensitive mechanoreceptors of the dactyl of the crab: Single unit responses during walking and evaluation of function. J. Neurophysiol. 57: 1601-1607
  • Loeb, J. (1918) Forced movements, tropisms and animal conduct. Lippincott, Philadelphia.
  • Massa, D., Ayers, J. and Crisman, J. (1992) Acoustic Communication, Navigation and Sensing Systems for a Biologically-based Controller for a Shallow Water Walking Machine, IEEE Proc.on Oceanic Systems, (1992) 590-595.
  • Maude, S. H. and Williams, D. D. (1983) Behavior of crayfish in water currents: Hydrodynamics of eight species with reference to their distribution patterns in Southern Ontario. Can. J. Fish. Aquat. Sci. 40: 68-77.
  • Pearson, K.G. 1972) Central programming and reflex control of walking in the cockroach. J. Exp. Biol. 56: 173-193.
  • Pinsker, H. M. and Ayers, J. (1983) Neuronal Oscillators. Chapter 9 in: The Clinical Neurosciences. Section Five. Neurobiology. W.
  • D. Willis [ed]. Churchill Livingstone Publishers. Pp. 203-266.
  • Sandeman, D. C. and Atwood, H. L. (1982) The Biology of Crustacea, Vol. 4. Neural Integration and Behavior. Academic Press, New York.
  • Selverston, A.I. and M. Moulins (1987) The Crustacean Stomatogastric System. Springer-Verlag. Berlin.
  • Sherrington, C. S. (1906) The Integrative Action of the Nervous System. Charles Scribner's Sons, New York.
  • Sillar, K.T., F. Clarac and B.M.H. Bush. (1987) Intersegmental coordination of central neural oscillators for rhythmic movements of the walking legs in crayfish, Pacifastacus lenuisculus. J. exp. Biol. 131: 245-264.
  • Stein, P.S.G. (1976) Mechanisms for interlimb phase control. In: Neural Control of Locomotion. Herman,RM; Grillner,S; Stein,PSG;
  • Stuart,DG [eds]. Plenun Publishing. New York. Pp. 465-487.
  • Stein, P.S.G. (1978) Motor Systems, with specific reference to the control of locomotion. Ann. Rev. Neurosci. 1: 61-81.
  • Wiersma, C.A.G. and Roach, J. L. M. (1977) Principles in the organization of invertebrate sensory systems.. In: Handbook of Physiology, Section 1: The Nervous System. Volume 1 Cellular Biology of Neurons. Brookhart, J. M. , Montcastle, V. V. B .
  • Kandel, E. M.[eds]. Williams and Watkins, Co. Baltimore. Pp. 1089-1087.
  • Wilson, D.M. (1967) Stepping patterns in tarantula spiders. J. exp. Biol. 47: 133-151.