TY - GEN
T1 - Emulation of biological motor primitives in an artificial system
T2 - Dynamic Systems and Control Division - 1999 (The ASME International Mechanical Engineering Congress and Exposition)
AU - Lee, Chae J.
AU - Reger, Bernard D.
AU - Tresch, Matthew C.
AU - Colgate, J. Edward
AU - Mussa-Ivaldi, Ferdinando A.
PY - 1999
Y1 - 1999
N2 - We have used observations of posture and movements in biological limbs to derive a controller for an artificial mechanism. The controller architecture emulates some of the known relations between spinal cord circuitry and the musculoskeletal system of vertebrates and, specifically, of the rat. This work relates to recent experiments suggesting that the neural circuitry of the spinal cord may be partitioned into a small set of functional modules. Activation of these modules, each connected to a set of limb muscles, resulted in force fields that have been measured at the endpoint of a limb. These force fields map each position of the foot into a corresponding static force vector. The force fields have been found to converge toward equilibrium positions located inside the leg's workspace. The experimental observation that vector fields induced by multiple stimulations add vectorially, suggested that convergent force fields form a system of building blocks (or 'primitives') for the generation of stable postures and movements. To emulate this biological mechanism in the control of an artificial two-joint limb, we established relationships among three hierarchical levels - spinal modules, muscles, and actuators - by deriving the mappings among the respective output fields. These mappings are used in combination with an inverse model of the actuators to calculate the actuator commands that generate a desired force field. We tested the ability of this control system to reproduce the force fields generated by the leg muscles of the rat and a set of force fields with significant geometrical features. Our results show that we can successfully and reliably transfer to our artificial system the features of muscle force fields. In addition, we exploited the same principle of vector summation observed in the biological system to combine these muscle fields into a variety of force field patterns, including the gradients of Gaussian potentials and locally parallel fields. We consider this a first step in the generation of a biomorphic motor control system. This work is supported by ONR grant N00014-95-1-0571 and NIH grant MH48185.
AB - We have used observations of posture and movements in biological limbs to derive a controller for an artificial mechanism. The controller architecture emulates some of the known relations between spinal cord circuitry and the musculoskeletal system of vertebrates and, specifically, of the rat. This work relates to recent experiments suggesting that the neural circuitry of the spinal cord may be partitioned into a small set of functional modules. Activation of these modules, each connected to a set of limb muscles, resulted in force fields that have been measured at the endpoint of a limb. These force fields map each position of the foot into a corresponding static force vector. The force fields have been found to converge toward equilibrium positions located inside the leg's workspace. The experimental observation that vector fields induced by multiple stimulations add vectorially, suggested that convergent force fields form a system of building blocks (or 'primitives') for the generation of stable postures and movements. To emulate this biological mechanism in the control of an artificial two-joint limb, we established relationships among three hierarchical levels - spinal modules, muscles, and actuators - by deriving the mappings among the respective output fields. These mappings are used in combination with an inverse model of the actuators to calculate the actuator commands that generate a desired force field. We tested the ability of this control system to reproduce the force fields generated by the leg muscles of the rat and a set of force fields with significant geometrical features. Our results show that we can successfully and reliably transfer to our artificial system the features of muscle force fields. In addition, we exploited the same principle of vector summation observed in the biological system to combine these muscle fields into a variety of force field patterns, including the gradients of Gaussian potentials and locally parallel fields. We consider this a first step in the generation of a biomorphic motor control system. This work is supported by ONR grant N00014-95-1-0571 and NIH grant MH48185.
UR - http://www.scopus.com/inward/record.url?scp=0033297725&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0033297725&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:0033297725
SN - 0791816346
T3 - American Society of Mechanical Engineers, Dynamic Systems and Control Division (Publication) DSC
SP - 897
EP - 905
BT - American Society of Mechanical Engineers, Dynamic Systems and Control Division (Publication) DSC
PB - ASME
Y2 - 14 November 1999 through 19 November 1999
ER -