288 pp. per issue
6 x 9, illustrated
2014 Impact factor:

Neural Computation

Summer 1989, Vol. 1, No. 2, Pages 173-183
(doi: 10.1162/neco.1989.1.2.173)
© 1989 Massachusetts Institute of Technology
New Models for Motor Control
Article PDF (593.05 KB)

What do a prototype robot (Brooks 1989) and a model for the control of behavioral choice in insects (Altman and Kien 1987a) have in common? And what do they share with a scratching cat (Shadmehr 1989)? The answer is distributed control systems that do not depend on a central command center for the execution of behavioral outputs. The first two in particular are examples of a growing trend to replace the long-held concept of linear hierarchical control of motor output with one of decentralized, distributed control, with inputs at many levels and the output a consensus of the activity in several centers.

Brooks (1989) describes a six-legged machine that, in its most advanced form, can walk over rough terrain and prowl around following a source of warmth, such as a person. The six legs, chosen as a compromise between stability and ease of coordination, give the robot a superficial resemblance to an insect — but the similarity goes deeper. The modular control system, designed strictly on engineering principles for maximum efficiency and economy, bears a striking similarity to the model we have proposed elsewhere (Altman and Kien 1987a) to describe the organization of the motor system in insects such as the locust. In both systems, the same set of components can generate different behaviors, depending on the context, and similar principles govern the generation of different levels of behavior, from movements of a single leg to coordinated responses of the whole beast. Neither requires a single center for integrating all sensory information and conflicts tend to be resolved by consensus at the motor level.