User:J. Lucas McKay/Proposed/Neuromechanical Modeling

From Scholarpedia
Jump to: navigation, search

  Dr. Lena H. Ting accepted the invitation on 16 September 2011

Neuromechanical modeling refers to computational or physical models of the interactions between the nervous system with the musculoskeletal systems and the physical environment in producing biological motion. Common motor behaviors such as walking, standing, playing sports, manipulating object, speech, and even digestion and breathing all arise from a complex interplay between the control of muscles from neural circuits, properties of musculotendon force generation, and interactions with the physical properties of the tissues and environments. Each of these underlying components is insufficient in predicting biological movement, as the interactions can lead to rather surprising complex emergent behaviors. Neuromechanical modeling provides a testbed for hypotheses about to role of neural control versus biomechanics in shaping biological movement.



Neuromechanical interactions govern how we move, but traditionally models for neural systems and biomechanical systems exist independently. However, because there is no one-to-on relationship between the impulse of a muscle and the resulting movement, it is not possible to infer motor behavior neural signals alone, and conversely, it is not possible to infer a neural control signal from biomechanical measures alone. This was first noted by Bernstein (1967) by examining the dynamic equations of motion of the musculoskeletal system based on Newtonian mechanics.

Hierarchical interactions (Ting 2009)

Purpose of neuromechanical models

To infer possible neural strategies for movement, their structure, their sensitivity, and their interactions with the biomechanics of the body and environment.

To interpret measured biomechanical and electromyographic (EMG) data.

Neuromechanical modeling integrates tools derived from several specialties, including rigid body dynamics, muscle models, and neural networks, to represent motor control of living organisms. Because performance of the nervous system is difficult to measure in isolation, neuromechanical modeling is often used to infer properties of the nervous system from more easily measured kinematics and kinetics. The neural and musculoskeletal systems are intimately connected, in part because neural commands act on the world only indirectly, through the muscles and bones, and in part because sensory feedback depends on the physics of the body's interaction with the environment.

To test neural control strategies accurately requires the the other aspect of the neuromechanical model - the muscles, the physics of the body, and any sensory feedback - be accurately described. Fortunately, these systems seems to be more amenable to testing in isolation. Muscle forces can be at least generally described by simple viscoelastic models, and body physics can be at least reasonably represented as a kinematic chain of rigid links. Accepting these models places the entire onus of matching experimental observations on the neural control model.

System capabilities predicted by neuromechanical models

Muscle coordination in static musculoskeletal models

Purpose: To investigate possible neural strategies for force generation

Static optimization and the force-sharing problem

  • Crowninshield

Feasible force set calculation

  • Valero-Cuevas + Kutch
  • McKay 2007, 2008

Motor control models supported by neuromechanical models

Muscle synergies

Raasch 1999 Bernicker 2009 Kargo and Giszter 2010 McGowan and Neptune 2010

Simple biomechanical feedback

Peterka 1999 Lockhart 2007 Scrivens robotic model Bingham Bunderson 2010

Optimal feedforward control

Franklin Sternad

Optimal feedback control

Bunderson 2007 Lockhart Neuromechanical modeling often incorporates elements of optimal control theory.

Proof of concept models

Neural circuitry models with mechanics

Ekeberg Rybak Cofer and Edwards 2010 Cabelguen, Ijspeert salamander model Loeb

Robotic approaches

Neurorobotics is essentially the physical implementation of neuromechanical modeling.

Chiel soft robotic aplysia model Ijspeert robotic salamander model Scrivens robotic postural model

Software packages




See also

Personal tools

Focal areas