This article presents a connectionist model of a central pattern generator for salamander locomotion. A 3D biomechanical simulation of the salamander’s body is developed whose muscle contraction is determined by the locomotion controller simulated as a leaky-integrator neural network. While the connectivity of the neural circuitry underlying locomotion in the salamander has not been decoded for the moment, the general organization of the designed neural circuit corresponds to that hypothesized by neurobiologists for the real animal. In particular, the locomotion controller is based on a body central pattern generator (CPG) corresponding to a lamprey-like swimming controller, and is extended with a limb CPG for controlling the salamander’s limbs. A genetic algorithm is used to instantiate synaptic weights of the connections within the limb CPG, and from the limb CPG to the body CPG, given a high level description of the desired gaits. A controller is thus developed which can produce a neural activity and locomotion gaits very similar to those observed in the real salamander. By varying the tonic excitation applied to the network, the speed, direction and type of gait can be varied. Movies of the simulations can be found at http://rana.usc.edu:8376/- ijspeert/ . Studying the locomotor circuitry underlying the locomotion of the salamander, an amphibian which presents both types of locomotion, may give some hints on the changes of the locomotor circuits which have accompanied this transition. The salamander makes axial movements during locomotion. It swims using an anguiliform swimming gait, in which the whole body participates to movement creation, and in which a wave of neural activity is propagated from head to tail, with an approximately constant wavelength along the spinal cord [5, 21. On ground, the salamander switches to a trotting gait, in which the body forms an S-shaped standing wave with the nodes at the girdles, which is coordinated with the movements of the limbs such as to increase their reach during the swing phase. EMG recordings have shown that two different motor programs underly these typical gaits, with a traveling of neural activity for swimming and a mainly standing wave during trotting [5, 21. The locomotor circuitry responsible for these motor programs has, however, not been decoded for the moment. It has been hypothesized that the salamander’s locomotor circuitry is based on a lamprey-like organization, with a lamprey-like central pattern generator for the body segments extended by a limb CPG for controlling the limbs [l, 21. In this article a central pattern generator is developed based on a similar assumption. This work is inspired by Ekeberg’s neuronal and mechanical model of the lamprey [4]. Similarly, a simple 3D mechanical simulation of a salamander in interaction with water or ground is developed whose muscular activity is determined by a simulated central pattern generator. The work presented here follows preliminary experiments on the control of a 2D salamander simulation [lo, 8, 91, and uses the same methodology as that used to develop potential swimming controllers for the lamprey [ll].
[1]
S. Grillner,et al.
Neural networks that co-ordinate locomotion and body orientation in lamprey
,
1995,
Trends in Neurosciences.
[2]
L. M. Frolich,et al.
KINEMATIC AND ELECTROMYOGRAPHIC ANALYSIS OF THE FUNCTIONAL ROLE OF THE BODY AXIS DURING TERRESTRIAL AND AQUATIC LOCOMOTION IN THE SALAMANDER AMBYSTOMA TIGRINUM
,
1992
.
[3]
J. Cabelguen,et al.
Fictive rhythmic motor patterns induced by NMDA in an in vitro brain stem-spinal cord preparation from an adult urodele.
,
1999,
Journal of neurophysiology.
[4]
T. Williams.
Phase coupling by synaptic spread in chains of coupled neuronal oscillators.
,
1992,
Science.
[5]
David Willshaw,et al.
From lampreys to salamanders: evolving neural controllers for swimming and walking
,
1998
.
[6]
Auke Jan Ijspeert,et al.
Synthetic Approaches to Neurobiology: Review and Case Study in the Control of Anguiliform Locomotion
,
1999,
ECAL.
[7]
J J Hopfield,et al.
Neurons with graded response have collective computational properties like those of two-state neurons.
,
1984,
Proceedings of the National Academy of Sciences of the United States of America.
[8]
John Hallam,et al.
Evolving Swimming Controllers for a Simulated Lamprey with Inspiration from Neurobiology
,
1999,
Adapt. Behav..
[9]
J. Cabelguen,et al.
Epaxial and limb muscle activity during swimming and terrestrial stepping in the adult newt, Pleurodeles waltl.
,
1997,
Journal of neurophysiology.