In this project, the PI from Washington University in St. Louis, Missouri and Co-PI from Texas Tech University in Lubbock, Texas propose to study the role of oscillatory patterns as an analog computational device. Specifically, the oscillatory patterns are generated by a spatially distributed array of weakly coupled dynamical systems, where the coupling is predominantly local, with perhaps a few long range interactions. Additionally, some of the dynamical systems in the network are driven by external inputs which modulate the spatiotemporal signal generated by the entire network.
The pattern generating circuit would be implemented in an array of MEMS (micro electro mechanical systems), in an array of coupled LASERS and in an array of coupled NEURONS. Each of the three arrays is well known for their ability to generate oscillatory patterns. For example, in the visual cortex of turtles, coupled neurons produce a propagating wave of activity that seems to respond to the visual input to the turtle retina. Like wise, an array of coupled LASERS seems to phase-lock into a sustained level of oscillation with a certain phase relation between two adjacent LASERS. Finally, an array of MEMS devices can be used to implement electro-statically controlled micro-actuators. These can be used to control locomotion of a land crawler or a swimmer.
For each of the three pattern generating circuits, the relationship between the controlled input and the generated oscillatory pattern would be studied with the following two specific tasks in mind. If the input is a sensory input, viz. visual input to the retina, the generated oscillatory patterns encode the sensory signals. For example, the location and velocity of a target is encoded in the spatiotemporal response of the generated cortical waves. An important task is to localize targets and to predict its future location by measuring the oscillatory patterns in the circuit. If the input to the circuit is a control input, viz. electrostatic input to a MEMS array, each element of the array is viewed as a signal generator that provides control input to a local actuator. An important task in locomotion is "gait control" which involves maintaining and switching between gaits. To achieve this task one needs to switch from one oscillatory pattern to another, perhaps as quickly as possible.
In summary, pattern generating circuits will be used for the purpose of sensory encoding and decoding with a specific aim towards "target tracking". They will also be used to provide coordinated control signals to a large number of actuators with a specific aim towards "locomotion control". An important extension of the proposed problem is "gait control" which involves smoothly switching between patterns in minimum time.
