EPFL LATSIS Symposium 2006 Dynamical principles for neuroscience and intelligent biomimetic devices

Two fundamental questions that can be asked about oscillatory activity are: how does it originate and what is its purpose? As will be shown, there are several ways such activity can be generated. It is not always clear what oscillatory function is even when the output leads directly to behavior. We have examined the dynamical properties of two small oscillatory systems, the lobster pyloric and gastric mill central pattern generators (CPGs). These CPGs are useful in providing answers to both questions because they are small, i.e. contain only a few neurons, and each neuron and synapse are repeatedly identifiable. The mechanisms and the infrastructure for producing the two different patterns have been well studied. Each neuron uses specific ionic conductances, particularly IH , IA, INaP and several ICa currents and synaptic connections with different characteristics to generate a three phase pyloric rhythm at 2 Hz and a six phase gastric mill rhythm at .1 Hz. The patterns can be turned on and off with specific input fibers and continuously regulated by sensory feedback. Modulators applied exogenously or by stimulation of identified neuromodulatory neurons can functionally “rewire” each CPG circuit to produce different stable patterns. What are the major take home messages ? • Some synaptically isolated pyloric neurons are chaotic. When the same neurons are synaptically reconnected, their activity becomes regularized. The purpose of this covert chaotic activity is not yet clear but may be to smooth the overall motor pattern as well as make it more adaptable. • Electronic neurons can replace biological neurons that have been inactivated thus helping to determine the role of each neuron in the overall operation of the circuit. Each neuron is unique and each plays a unique role. • Pyloric CPGs from different animals connected with a dynamic clamp show how best to couple separate unit oscillators. Reciprocal inhibitory connections connected to driver cells can produce stable out-of-phase patterns. The same connections to particular non-driver cells produce stable in-phase patterns. Changing the strengths of the inhibitory synapses produce intermediate phase relationships between the two CPGs. • Sensory inputs provide feedback to both CPGs and assist in producing effective and smooth oscillatory behavior as well as dealing with cycle-by -cycle modifications due to sensory inputs. • The core circuits, i.e. the minimal network that can generate the pattern, for both the gastric mill and the pyloric rhythm can be modeled and may be best described with a winnerless competition algorithm made up of asymmetric inhibitory synapses. A model CPG circuit, based on principles learned from the stomatogastric system and implemented in analog hardware can act as a controller for a robotic lobster leg. By using presynaptic inhibition and sensory feedback, this small network can provide a closed loop microcontroller for forward and backward stepping.