Meeting Abstract
Hexapod animal models, such as cockroaches and stick insects, have proven to be very useful in the development of biomimetic robots. To understand hexapod neural control and apply it to robots, we first need to understand both the biomechanics of insect limbs and the interplay between neural activity and limb actuation. We will investigate this problem in Drosophila because our knowledge of its genetics make it possible to investigate single leg control in completely intact animals. We began our investigation by estimating the passive properties of the limb: all motor neurons in tethered flies were optogenetically deactivated into a ‘passive’ state where the only forces present are those produced by the tissue viscoelasticity and gravity. We use these passive kinematics to compute passive muscle torques and determine the fly’s joint tissue material properties, namely the torsional modulus and damping constants. Our experiments reveal that the cuticle and passive muscles at the Coxa-Femur and Femur-Tibia joints function as linear angular springs with some viscosity. Next, we model the 3D kinematics of the fly’s legs using an actuated damped spherical double pendulum which we derive using Lagrangian mechanics. This model accounts for all of the torques present in the passive case, and actuation is provided by a spiking neuronal network which modulates the stiffness of springs in the joints of the pendulum model. We fit this model to 3D data from actively moving flies suspended from a tether. Finally, we describe how the model recapitulates neural control of fly leg kinematics using mutual information and transfer entropy, which describe correlation and informational flow between observable signals, respectively.
