Hard-wired Central Pattern Generator Hardware Network for Quadrupedal Locomotion Based on Neuron and Synapse Models

Abstract

Many different quadrupedal walking patterns (gaits), such as walking , trotting , bounding and galloping can be generated by systems of coupled central pattern generators (CPG). However, the physiological mechanisms for the walking patterns are unclear. As a result, from an engineering viewpoint, many different mathematical models have been proposed to describe these walking patterns. In this report, we propose a hard-wired CPG network based on Rybak s model that can reproduce quadrupedal locomotion walking patterns. In this network, we use the beating model proposed by Hoshimiya et al. and the bursting model proposed by Maeda and Makino. The main purpose of this study was to reproduce the typical walking patterns; walking and bounding , with a hardware model, and to switch between these patterns using only one parameter, which can be interpreted as voltage stimulation from the midbrain locomotor region. We found the transition from the walking to the bounding behavior to be due to a relative weakening of the coupling between the CPGs in the network by stimulation from the midbrain locomotor region.

Ultra-low-power switching circuits based on a binary pattern generator with spiking neurons

“…So far, several central pattern generators have been artificially fabricated in the form of electronic circuits for controlling robot actuators 13 – 17 , biological muscles 18 , and physiological systems 19 . However, there are few examples of their application to controlling more general electronic circuits, such as the switching circuits (Fig.…”

“…Artificial central pattern generators have been previously studied in the field of robotics 13 – 17 . There, the information is conveyed as the analog value of the spike rate (spike rate coding), which can be used to control the actuators of robots.…”

Current‐input current‐output analog half center oscillator and central pattern generator circuits with memristors

“…One of the most important problems to address is switching between different gait modes without having to change the circuit topology completely. There are examples of hard‐wired CPG networks (using bipolar junction transistors (BJTs)) which can switch between walking and bounding modes by the use of a single control parameter (without having to alter circuit topology) . Numerical methods can simulate behavior of CPGs easier in terms of achieving smooth transition between different gaits and learning new locomotion patterns.…”

“…There are examples of hard-wired CPG networks (using bipolar junction transistors (BJTs)) which can switch between walking and bounding modes by the use of a single control parameter (without having to alter circuit topology). 6 Numerical methods can simulate behavior of CPGs easier in terms of achieving smooth transition between different gaits and learning new locomotion patterns. One example of such numerical methods has been used by a robotic fish.…”

“…The difference between the 2 types of connections simply lies in the type of current mirror used, ie, p-type current mirrors result in − ("inhibitory"), n-type current mirrors lead to + ("excitatory"), where represents synaptic strength. For example, in the CPG operating in walk mode, limb RF is connected to limb LH with a p-type current mirror, drawing I RF u {2,1} from tonic input I s , see Equation 6. All the other connections between limbs are made using n-type current mirrors, see Equations 5, 7, and 8.…”