10.16.2009 - Advances in Computational Motor Control

2009 Advances in Computational Motor Control

Symposium at the Society for Neuroscience Meeting

Friday, October 16, 2009, 1:00 PM - 7:40 PM

Chicago Convention Center, Room S504ABCD

1:00 - 1:05 Opening Remarks

Session 1: platform presentations

Invited talk: Michael Graziano (Princeton University)

Mapping behavioral repertoire onto the motor cortex

    • Copied stimulation patterns usually used in examining eye movements
    • Stimulating motor cortex for a period of 1 sec
      • Stimulate complex behaviors
        • Hand to mouth
        • Reaching movement
        • "Manipulating movement"
        • Bilateral movements
        • Climbing movements
      • Behavioral repertoire is mapped out in "lumps" of motor cortex
      • Behavioral map follows the somatotopic map
      • "Spatial map" of hand space is mapped out in different regions of the brain
    • Different possible maps
      • Somatotopic map
      • Behavioral map
      • Spatial map corresponding to action space
    • How do you map high dimensions to 2D cortical sheet? <-Dimensional reduction is necessary?!!
    • Hypothesis: Dimensional reduction of information using an optimal "local continuity" of dimensional reduction into 2D space
    • Parameterization of body movements (No kinematics...interesting)
      • Body parts
      • Hand locations
      • Common types of behavior
    • The dimensional reduction method was the ?Cajonan? algorithm
      • Starting with a random seed produces results that are transformed affinely
      • Used a basic seed of a somatotopic map
    • There is a match between the somatotopic map seen in physiology and the optimal "mapping" of efficient grouping of movement behaviors onto a 2D manifold.
      • This efficient mapping matches:
        • Somatotopic map
        • Supplementary motor cortex map
    • The fundamental truth of cortical organization is not one of "maps": It is one of trying to minimize distance between like behavior.

Sara Steenrod and Michael Goldberg (Columbia University)

Environmental memory in parietal cortex: A possible substrate for guidance of movement in the absence of visual stimulation

    • Use the parietal cortex when navigating the dark (spatial mapping?)
    • Looking in the intraparietal area (LIP)
      • A place to store environmental memory
    • Recording single neurons in monkeys
      • Fixation-saccade task with a distraction
    • Most cells in the LIP had "memory" of the distracting stimulus, but this fades away with time.
      • It lasts for about 10 trials.
      • I imagine that this phemonenon is conserved across all neural structures and this is probably happening when we are talking about "gains" that are being tuned. Maybe we need to look at some of these papers when we start getting some data from Julia's studies.
    • There is a speculation that the LIP has an egocentric map of the environment.
      • Could be through proprioceptive mapping.
      • Questions about whether or not this is due to a retinal "after-image" storing the memory.

James Rebesco, Sara Solla and Lee Miller (Northwestern University)

Rewiring neural connectivity by micro-stimulation

    • Examining plasticity for the motivation of a bi-directional brain-machine interface
    • Difficulties in creating an interface:
      • Small population that can be recorded from
      • The connectivity of those neurons is unknown
    • Functional connectivity algorithm is determined based on spike firing rate as estimated by the possible connections that might occur between the observed neurons and all other connections
  • What about if the connectivity/weighting changes while you are observing?
    • They were unable to determine the connectivity, but felt like they were able to identify changes in the connectivity of the neuron.
    • They use spike timed stimulation to try and reproduce Hebbian learning.
      • Saw changes in weighting if the spike timed potentiation was ~5 ms

Maarten Frens, Beerend Winkelman and Opher Donchin (Ben Gurion University)

Forward models and state estimation in compensatory eye movements

    • Rabbit does not have a fovea! This behavior is cool because:
      • Minimize retinal slip
        • Optokinetic rflex
          • Low frequency sensitivity (closed loop)
          • It has a delay of 100 ms <-- DESTABILIZING!
            • Needs a compensatory model for prediction
        • Vestibulo-ocular reflex
          • High frequency sensitivity (open loop)
      • Cerebellar involvement - flocculus
      • No cortical movement
    • Look at "Shadmehr and Krakaur 2009" hopefully to look at loop diagram
    • Flocculus: Motor command VS Prediction command
      • Motor command should see a delay with eye movement because the relationship is causal
      • Predictor command should be have the spiking aligned with the eye movement
    • The motor command and the kinematics are uncorrelated instantaneously. <--Just like us! Maybe Seyed should be doing the same thing for the random platform movements.

Session 2: platform presentations

Julian Tramper, Bert Kappen, and Stan Gielen (Radboud University Nijmegen)

Predicting human motor performance using stochastic optimal control

    • Small variability make decision now
    • Large variability make decision later
    • Optimal control
      • optimal cost-to-go via Bellman optimality
      • Hamilton Jacobian Bellman equation gives a mathematical formalism to the first two statements about variability.
    • What happens if noise is state dependent?
    • What happens if there is observation noise?
    • Does the task constrain the output?

Miriam Zacksenhouse, Koren Beiser, Joseph O'Doherty, Mikhail Lebedev and Miguel Nicolelis (Technion and Duke University)

Optimal control framework successfully explains changes in neural modulations during experiments with Brain Machine Interfaces

    • Looking at tuning curves of the endpoint velocity in relation to neuron firing rate

Gary Sing, Simon Orozco and Maurice Smith (Harvard University)

Adaptive responses in the human motor system interpret arbitrary force perturbations as state-dependent dynamics

    • Position, velocity dynamics can be easier/harder to learn.
      • Actions through feedback presume state dependence
      • Forces must necessarily go through a mechanical system that acts as a filter
      • Anisotropy in motor learning of force fields appears to be dependent on velocity and position.

Jörn Diedrichsen, Niall Lally and Ian O’Sullivan (UCL and Bangor University)

When two systems work as one: Minimizing signal-dependent noise through Nash-Equilibria

    • Minimize u^2 (energy/noise minimization)
    • Need to include both effort and variability in cost function
    • To optimize "laziness" need to find a "nash equilibrium" across the muscles
    • Suggests heirarchical control vs coordinated central control

Posters

Gary Sing, Bijan Najafi, Adenike Adewuyi and Maurice Smith (Harvard University)

A novel mechanism for the spacing effect: Competitive inhibition between adaptive processes can explain the increase in motor skill retention associated with prolonged inter-trial spacing

S. Schaefer, I. Shelly and Kurt Thoroughman (Washington University)

Beside the point: Motor adaptation in task-irrelevant conditions

Jordan Taylor, Azeen Ghorayshi and Richard Ivry (UC Berkeley)

The Cost of Strategic Control: Attenuation of Adaptation

Jasper Schuurmans, Winfred Mugge, Alfred Schouten and Frans van der Helm (University of Tewnte)

Sensory weighting of force and position feedback in human motor control tasks

Jonathan Dingwell, Joby John and Joseph Cusumano (UT Austin)

Computational models of goal equivalent control in human treadmill walking

Jun Izawa and Reza Shadmehr (Johns Hopkins University)

The disparate roles of reward and sensory prediction errors in learning motor control

Ian Stevenson, Hugo Fernandes, Iris Vilares, Kunlin Wei and Konrad Körding (Northwestern University)

Bayesian integration and non-linear feedback control in a full-body motor task

    • Look up Kording (2007) Science 318

Sandro Mussa-Ivaldi, M. Casadio and A. Pressman (Northwestern University)

Adaptive force control

Lucas McKay and Lena Ting (Georgia Tech)

The nervous system maps high-dimension sensory inflow to low-dimension motor outputs during postural responses

Natalia Dounskaia, Keith Nogueira and Elizabeth Drummon (Arizona State University)

Hierarchical control of bimanual movements revealed by arm dominance challenges the muscle homology principle

Session 3: platform presentations

Invited talk: Zoran Popovic (University of Washington)

Towards high-fidelity high-dimensional natural controllers

    • Evolution in fast-forward
      • Continuous optimization of shape and form
      • Solve for pose and actuators
    • Optimization through global minimization
      • NEED TO LOOK THIS UP
    • Want to determine optimal cost function based on a given set of input output data
      • This is Karen Liu's work
    • Controller for simulation
      • The input states include the joint angles AND ground contact forces

Vishwanathan Mohan, Pietro Morasso, Giorgio Metta and Jacopo Zenzeri (Italian Institute of Technology)

Equilibrium point hypothesis revisited: Advances in the computational framework of Passive Motion Paradigm

    • Didn't show

Tom Erez (Washington University)

An interpretive model of hand-eye coordination

    • In visual search, eye saccades to "center" of the scene
    • Saccades in the model occurred to disambiguate the obstacles
    • In the model the eye stared at the hands, but physiologically you usually look at the target.

Giby Raphael, George Tsianos and Gerald Loeb (University of Southern California)

Spinal-like regulator for controlling wrist movements

    • Many different gains are good enough to meet kinematic task
    • This mirrors a similar idea by (Prinz 2004) that variability in input space does not necessarily mean variability in output space.