Pulse-type Hardware Neural Networks Research Articles

The walking and running control of a human musculoskeletal model using a low-power consumption hardware central pattern generator model

Applying a control system with low energy consumption and low load of motion control to robots, similar to living organisms, is considered to be one of the most important issues in robot development. We have been studying systems that use pulse-type hardware neural networks to control robotic motion with a small number of control signals, as is the case in living organisms. In particular, it has been mimicking the function of the central pattern generator localized in the spinal cord of living organisms to generate motion patterns. In the present article, a new biomimetic control system using pulse-type hardware neural networks for biped gait control is reported.

Biped Control of the Musculoskeletal Model with Pulse-Type Hardware Neural Networks

Biped Control of the Musculoskeletal Model with Pulse-Type Hardware Neural Networks

Spike-timing-dependent plasticity model for low-frequency pulse waveform

The authors are studying pulse-type hardware neural networks (P-HNN) for generating the driving waveform of the robot. The P-HNN generates a driving waveform by the oscillatory pulse waveform. Oscillating frequency depends on the time constant of the oscillator. In the case of implementing into the integrated circuit (IC), a large time constant cannot construct to the limited space. This paper is discussing the generation of low-frequency pulse waveform without using a large time constant. Two cell body models pre-synaptic neuron (Pre) and post-synaptic neuron (Post) generate the high-frequency pulse (637 kHz). A slight time difference of both pulse waveform converts to the small voltage difference of spike-timing-dependent plasticity (STDP) model. The small voltage difference indicates the coupling coefficient of Pre and Post. The value of the coupling coefficient changing slowly, thus, the output neuron (Out) generates the burst-like waveform which can use as a low-frequency pulse waveform (37 Hz).

Hardware neural network models of CPG and PWM for controlling servomotor system in quadruped robot

This paper discusses the pulse-type hardware neural networks (P-HNNs) that contain a central pattern generator (CPG) and a pulsewidth modulation (PWM) servomotor controller and the application to quadruped robots. The purpose of our study is mimicking the biological neural networks and reproducing the similar motion of the living organisms in the robot. The CPG of the living organism generates the walking rhythms. We mimicked this CPG by modeling the cell body and the synapse of the living organism. The developed CPG composed of the P-HNN output four pulse signal sequences and the four outputs are introduced to each leg of the quadruped robot. On the other hand, the angle of the servomotor is controlled by the PWM. The PWM is obtained by modeling the axon of the living organism. The CPG and the PWM servo control system perform the walking motion of the quadruped robot. Moreover, the gate pattern change of quadruped animals is reproduced by these P-HNNs.

Gait pattern changing of quadruped robot using pulse-type hardware neural networks

This paper studied about gait pattern changing of the constructed quadruped robot system using pulse-type hardware neural networks (P-HNN). We constructed the 20 cm in size prototype quadruped robot system. Quadruped robot system consisted of mechanical components and electrical components. The mechanical components consisted of four legs, body frames and four servo motors. Quadruped animal-like locomotion could realize by only four servo motors using link mechanisms to each leg. The electrical components consisted of P-HNN, power supply circuit, control board and battery. P-HNN was constructed by analog discrete circuits which could mount on top of the quadruped robot. As a result, constructed P-HNN could output the locomotion rhythms which were necessary to generate the gait pattern of the quadruped robot. P-HNN could output the locomotion rhythms without using software programs or analog digital converter. In addition, P-HNN could change the locomotion rhythms by inputting the trigger pulse to the P-HNN. Our constructed quadruped robot system could perform the locomotion without using external devices.

Development of quadruped robot with locomotion rhythm generator using pulse-type hardware neural networks

This paper discussed about development of quadruped robot which could perform the quadruped animal-like locomotion. Locomotion rhythm of the quadruped robot was generated using the pulse-type hardware neural networks (P-HNN). Quadruped robot had mechanical components and electrical components. The mechanical components of the quadruped robot consist of the body frame, link mechanisms, 4 legs and 4 servo motors to realize the quadruped animal-like locomotion. The body frame, link mechanisms and 4 legs were made from aluminum base alloy. The electrical components of the quadruped robot consist of control board, battery and P-HNN. P-HNN generates the locomotion rhythms using synchronization phenomena such as biological neural networks. The control board actuates the servo motors according to the generated locomotion rhythms. As a result, constructed quadruped robot could perform the quadruped animal-like locomotion using the generated locomotion rhythm, which was shown in this paper.

Locomotion Control of MEMS Microrobot Using Pulse‐Type Hardware Neural Networks

SUMMARYThis paper presents the locomotion control of a microelectromechanical system (MEMS) microrobot. The MEMS microrobot demonstrates locomotion control by pulse‐type hardware neural networks (P‐HNN). P‐HNN generate oscillatory patterns of electrical activity like those of living organisms. The basic component of P‐HNN is a pulse‐type hardware neuron model (P‐HNM). The P‐HNM has the same basic features as biological neurons, such as the threshold, the refractory period, and spatiotemporal summation characteristics, and allows the generation of continuous action potentials. P‐HNN has been constructed with MOSFETs and can be integrated by CMOS technology. Like living organisms, P‐HNN has realized robot control without using software programs or A/D converters. The size of the microrobot fabricated by MEMS technology was 4 × 4 × 3.5 mm. The frame of the robot was made of a silicon wafer, equipped with rotary actuators, link mechanisms, and six legs. The MEMS microrobot emulated the locomotion method and the neural networks of an insect by rotary actuators, link mechanisms, and the P‐HNN. We show that the P‐HNN can control the forward and backward locomotion of the fabricated MEMS microrobot, and that it is possible to switch its direction by inputting an external trigger pulse. The locomotion speed was 19.5 mm/min and the step size was 1.3 mm. © 2013 Wiley Periodicals, Inc. Electr Eng Jpn, 186(3): 43–50, 2014; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/eej.22473

1P1-B09 生物の脳を模したパルス形ハードウェアニューラルネットワークによるサーボモータ用PWM波形の生成(脳・神経・認知ロボティクス)

This paper presents the pulse-type hardware neural networks (P-HNN) which could output the pulse width modulation (PWM) waveform of PWM servo motor for robots. Basic components of P-HNN were pulse type hardware neuron model (P-HNM). P-HNM consisted by the cell body model, the axon model and the synaptic model. P-HNM had the same basic features of biological neurons such as threshold, refractory period, spatio-temporal summation characteristics and enabled the generation of continuous action potentials. We used the synaptic model which could vary the synaptic weights by controlling voltage. As a result, we showed that P-HNN could control the PWM servo motor from 0 to 180 degrees.

Locomotion Control of MEMS Micro Robot Using Pulse-Type Hardware Neural Networks

This paper presents the locomotion control of micro electro mechanical systems (MEMS) micro robot. MEMS micro robot demonstrates the locomotion control by the pulse-type hardware neural networks (P-HNN). P-HNN generates oscillatory patterns of electrical activity such as living organisms. The basic component of P-HNN is pulse-type hardware neuron model (P-HNM). P-HNM has same basic features of biological neurons such as threshold, refractory period, spatio-temporal summation characteristics and enables the generation of continuous action potentials. P-HNN was constructed by MOSFETs, can be integrated by CMOS technology. Same as the living organisms P-HNN realized the robot control without using software programs, or A/D converters. The size of micro robot fabricated by the MEMS technology was 4×4×3.5 [mm]. The frame of robot was made of silicon wafer, equipped with rotary type actuators, link mechanisms and 6 legs. MEMS micro robot emulated the locomotion method and the neural networks of the insect by rotary actuators, link mechanisms and P-HNN. As a result, we show that P-HNN can control the forward and backward locomotion of fabricated MEMS micro robot, and also switched the direction by inputting the external trigger pulse. The locomotion speed was 19.5 [mm/min] and the step width was 1.3 [mm].

Pulse-type hardware neural networks for olfactory center model of land slug

Pulse-type hardware neural networks for olfactory center model of land slug

四足歩行運動パターンを生成・移行可能なパルス形ハードウェアCPGモデル

The purpose of our research is to clarify information processing functions of living organisms by neural networks using pulse-type hardware neuron models and applying pulse-type hardware neural networks to engineered models. It is known that locomotion such as walking by a living organism is generated and transited by CPG (Central Pattern Generator) in the central nervous system. We investigate a pulse-type hardware CPG model using coupled oscillator composed of pulse-type hardware neuron models. A CPG model is need to generate and control quadruped locomotion. In this paper, we describe generation and transition of oscillation patterns, corresponding to quadruped locomotion patterns. As a result, it is shown that generation and transition of oscillation patterns are possible by giving external inputs of one pulse to the CPG model.