Biological motor controlSlide 2Motor neuroscience: levels and themesSlide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Biological motor controlAndrew RichardsonMcGovern Institute for Brain ResearchMarch 14, 2006Why bother with biology?Both plant (musculoskeletal system) and controller (nervous system) optimized by evolution for versatility and efficiency.Motor neuroscience: levels and themesMechanics of the neuromusculoskeletal systemSimplification of control by neural modulation of the mechanicsMotor behaviorModel-based adaptive control in changing environmentsCoordinate frames and methods for movement planningMotor neurophysiologyNeural representations in motor cortexCortically-controlled neural prostheticsMechanics of the neuromusculoskeletal systemvelocitytensionlengthening shorteningmuscle lengthtensionForce production by a muscle is dependent on its length and velocity.Mechanics of the neuromusculoskeletal systemChanges in muscle activation cause changes in the viscoelastic properties of the muscle. muscle lengthtensionThe nervous system receives information on muscle length, velocity, and force.golgi tendon organmuscle spindleMechanics of the neuromusculoskeletal systemReflexes are local feedback loops that can modify the viscoelastic behavior of the motor periphery.Mechanics of the neuromusculoskeletal systemReflex gains can be modulated by higher levels of the neural controller (e.g. cerebellum, motor cortex).Nichols and Houk, 1976Thus, total neuromuscular viscoelasticity has both intrinsic (i.e. muscle) and reflexive contributions. The gain of each of these contributions can be modulated by central commands.At least for some behaviors, the neural controller likely takes advantage of these mechanical properties …1. Equilibrium-point (servo) control2. Impedance controlMechanics of the neuromusculoskeletal systemEquilibrium-point controlEquilibrium points dependent on neuromuscular elasticity & loads.Changing the stiffness ratio for antagonistic muscles shifts the equilibrium point (producing a “virtual trajectory”). Precludes need to compute inverse dynamics, thus simplifying neural computations. Mechanics of the neuromusculoskeletal systemflexor tensionextensor tensionjoint angleBizzi et al, 1984Mechanics of the neuromusculoskeletal systemImpedance controlIn addition to viscoelastic behavior, inertial behavior can be modulated due to the kinematically redundant skeleton.Modulate full mechanical impedance (force-length, force-velocity, and force-acceleration relationships) of the limb for improved stability or performance.Burdet et al, 2001Hogan, 1985Motor neuroscience: levels and themesMechanics of the neuromusculoskeletal systemSimplification of control by neural modulation of the mechanicsMotor behaviorModel-based adaptive control in changing environmentsCoordinate frames and methods for movement planningMotor neurophysiologyNeural representations in motor cortexCortically-controlled neural prostheticsMotor behaviorAdaptive control of reaching movements generally leads to proactive, not reactive (i.e. impedance control), compensation.B = 0 B = 0B =Motor behaviorLackner and DiZio, 1994Motor behaviorControl of reaching movements may depend on the type of task (e.g. nominally stable or unstable) and familiarity with the task.Osu et al, 2003But how do we plan motions?Motor behaviorMotions are planned in endpoint coordinates. Morasso, 1981Endpoint hand paths are straightEndpoint velocityis “bell-shaped”Joint trajectories are complexMotor behaviorOptimization criteria have been proposed to explain the observed behavior, given the seemingly ill-posed problem of getting from point A to point B. Minimum jerk (Hogan, 1984)Minimum torque change (Uno et al, 1987)Minimum endpoint variance (Harris and Wolpert, 1998)Endpoint hand paths are straightEndpoint velocity is “bell-shaped”Motor behaviorBut production of a fixed “desired trajectory”, whether through optimization or not, does no account for some features of motor behavior.For example, it doesn’t predict increased variability in dimensions of state-space that are not relevant to the task. Todorov et al, 2005Todorov, 2004Motor behaviorOptimal feedback control may provide a better description of motor planning, as well as motor execution.But what is the cost function?Motor neuroscience: levels and themesMechanics of the neuromusculoskeletal systemSimplification of control by neural modulation of the mechanicsMotor behaviorModel-based adaptive control in changing environmentsCoordinate frames and methods for movement planningMotor neurophysiologyNeural representations in motor cortexCortically-controlled neural prostheticsMotor neurophysiologyMotor neurophysiologyMotor neurophysiologyMotor neurophysiologyWhat is encoded in motor cortex – kinematics or dynamics?Neural recording experiments Kakei et al, 1999What is encoded in motor cortex – kinematics or dynamics?Neural stimulation experiments Graziano, 2002Motor neurophysiologyNicolelis, 2001Motor neurophysiologyEngineering application: cortically-controlled neural prostheticsMotor neurophysiologyNeural control of a robot Wessberg et al,